Utilitas Mathematica

DOMINATION ON SOME SPECIAL GRAPH

2025-08-11T06:44:45+00:00

A set of vertices ???? is said to dominate the graph

Capturing Spatio-Temporal Patterns for Intrusion Detection: A Hybrid CNN-LSTM-GRU Model on the NSL-KDD Dataset

2025-08-12T13:39:47+00:00

Intrusion Detection Systems (IDS) are essential for detecting and preventing unauthorized activities in computer networks. This research introduces a DL based IDS framework using the NSL-KDD dataset, employing advanced architectures such as GRU, LSTM, CNN, and a hybrid CNN-LSTM-GRU model. The system addresses both binary and multi-class classification tasks to distinguish between normal and malicious traffic, as well as identify specific attack categories like DoS, Probe, R2L, and U2R. Among the evaluated models, the hybrid CNN-LSTM-GRU approach achieved superior performance due to its ability to capture both spatial and temporal patterns in network data. The results demonstrate that deep learning (DL) significantly enhances the accuracy and robustness of intrusion detection, offering a scalable and intelligent solution for network security.

Multi-Domain Seismic Noise Reduction in 2D Land Data: A Case Study from Poland

2025-08-13T00:54:49+00:00

Seismic reflection data are often contaminated by various types of noise, particularly ground-roll, which can obscure true reflection signals and hinder data interpretation. Effective noise reduction is essential for improving the signal-to-noise ratio and revealing subsurface structures. Many noise suppression techniques rely on transforming seismic data into different domains, making it easier to separate noise from useful signals. In this study, we applied a range of algorithms across multiple domains—including shot, receiver, t-x, f-x, f-k, R-T, and offset class domains to effectively reduce both coherent and random noise in 2D seismic land data. Our results demonstrate that these multi-domain filtering methods significantly enhance seismic image quality, enabling clearer and more reliable structural and stratigraphic analysis.

Investigating the reliability of using smartphone app and smartphone sensors in undergraduate laboratory experiments

2025-08-18T19:06:13+00:00

In the present work we investigated a simple and interesting experiment in physics- Dopler effect- using smartphones to study the reliability of using the smartphone app ‘Phyphox’ as a laboratory tool to perform various experiments in Physics. Doppler effect is a fundamental principle in wave mechanics with numerous applications in various scientific fields. While our experiment modelled Doppler redshift with sound, it helps us visualize why astronomers observe redshifted light from galaxies. Though cosmological redshift and Doppler shift differ in their underlying physics, the understanding of Doppler shift plays a very significant role in the development and understanding of cosmological red shift, that is a key piece of observational evidence supporting the Big Bang theory. In our experimental setup, two smartphones and a laptop are taken to study the Doppler effect. One of the smartphones is mounted on a stand which acts as a stationary source of monochromatic sound and the other, with the app ‘Phyphox’ installed acts as a moving observer. A laptop is connected to the second phone to record the frequency change as the observer moves. The experimental results of observed Doppler effect are quantitatively compared with the theoretical predictions. The agreement of experimental data with the theoretical data indicates that the method adopted is valid and the app used is reliable.

Selection of Comparable Subjects from Different Treatment Groups When Randomization is Not Feasible

2025-08-20T08:40:33+00:00

Objective: The aim of this research is to find solutions for selecting comparable subjects in non-randomized situations across different treatments.
Method: A series of analysis were performed to find subjects with similar profiles. The Propensity Score (PS) was estimated using logistic regression with potential covariates and a Cartesian product of PS across the subjects and treatments was created. The absolute difference in PS was derived and the pair with the minimum difference in PS was obtained, finally dose effect was estimated from selected paired.
Results: There were 23 subjects in the low dose group and 186 in the high dose group under treatment to assess the effectiveness of the ADAS-cog score in terms of change from Baseline to Week 24. The unequal proportion of subjects between the two dose groups has raised questions about balance and bias. An optimum matching technique was used to find profiles similar to the 23 subjects in the high dose group. Although the result was non-significant, it was more reliable because it was based on similar profiles, inadequate sample size could be an issue. The re-sampling technique was utilized to generate 80 more samples from the original 23 subject data from low dose. The dose effect was again estimated and found to be significant at Week 16 and Week 24.
Conclusion: The PS method was able to find subjects with similar profiles across the dose groups and provide reliable results regarding dose differences. After re-sampling and making covariate adjustments, the dose difference was found significant.

Integrability and periodic orbits in the generalized quasispecies model

2025-08-20T08:42:59+00:00

This paper studies a parametric family of systems of differential equations, which is obtained from the quasispecies model assuming arbitrary parameters without biological constraints. We study equilibria, invariant manifolds and the integrability of the family. It is well known that, by restricting the parameters to the biological domain, the quasispecies model is integrable, and here, we show that it is also the case for general parameters. Moreover, we consider a 4-dimensional realization of the model under the influence of a periodic perturbation. After restricting the system to an invariant manifold and relying on the first-order averaging technique, we demonstrate the existence of unstable periodic orbits in a neighborhood of the equilibrium located at the origin. It shows that periodic orbits emanate from the Zero-Hopf bifurcation that the mentioned equilibrium undergoes when the small parameter equals zero.

Exploring the Key Factors Behind Overweight and Obesity: Insights from BMI Data

2025-08-22T16:36:52+00:00

Leisure is commonly understood as free time beyond work and daily responsibilities, or as a quality of experience marked by personal freedom, enjoyment, and voluntary engagement. Globally, the rising prevalence of overweight and obesity among adults has been strongly linked to unhealthy eating habits and physical inactivity, with sedentary lifestyles becoming more common due to modern work routines and technology use. This study explores the relationship between leisure time activities and overweight or obesity among adults in Kozhikode Corporation, using population-based surveys to analyze lifestyle patterns. Activities such as dancing, aerobics, Zumba, and morning walks are recognized for their positive impact on physical and mental health, and many individuals consider structured workouts like yoga, gym sessions, or jogging as part of their leisure routine. By examining these habits within the local context, the study aims to highlight how active leisure can contribute to weight management and inform effective community-based health promotion strategies.

Detection of Mental Stress in Sports University Students through Machine Learning Techniques

2025-08-26T15:26:31+00:00

Mental stress is a significant issue affecting young individuals, particularly students, and can lead to serious cognitive disabilities, if left unaddressed. This study aimed to develop a machine-learning-based system to detect and measure stress levels in university-level sports students using vocal and acoustic features. Data were collected from 2400 students at the Lakshmibai National Institute of Physical Education (NERC), Guwahati, and analysed using Convolutional Neural Network (CNN) and random forest (RF) classification algorithms. The impact of exam pressure, match pressure, and recruitment stress on mental stress levels was examined. The performance of the algorithms was evaluated using the accuracy, precision, recall, and F1-score metrics. The RF algorithm achieved the highest accuracy (91.1 %) among the two classifiers. The proposed system aims to provide an objective tool for assessing stress levels, enabling earlier intervention, and more effective management of stress-related conditions by clinical psychologists. The study hypothesized that certain vocal characteristics, such as pitch variability, energy, Mel Frequency Cepstral Coefficients (MFCCs), Linear Predictive Coding Coefficients (LPCC), Zero Crossing Rate (ZCR), formant extraction, tempo beat extraction, and tonnetz extraction, would exhibit a significant correlation with higher stress levels. This study reviewed research on stress detection with machine learning, summarizing methods and classifier performance.

GRAPH-THEORETIC AND STATISTICAL MODELS FOR DETECTING AND TRACING DEEPFAKE MEDIA IN INTERCULTURAL COMMUNICATION NETWORKS

2025-08-28T05:34:04+00:00

This study investigates the detection and tracing of deepfake media in intercultural communication networks through the integration of graph-theoretic modeling and statistical analysis. Drawing on empirical data and secondary scholarship, it demonstrates how detection accuracy varies across formats such as movie scripts, press releases, social media posts, and email campaigns. Findings reveal that detection outcomes are shaped not only by algorithmic precision but also by cultural trust structures and network resilience. Western clusters, characterized by decentralized communication patterns, exhibit stronger resistance to synthetic content, while hierarchical clusters show heightened vulnerability. By coupling mathematical modeling with cultural analysis, the study contributes a replicable methodology and offers insights for both academic and practical applications. It argues that effective deepfake detection requires the alignment of technical innovation with culturally responsive strategies.

TECHNOLOGY-DRIVEN ENTREPRENEURSHIP: EXPLORING THE RISE OF TECH STARTUPS IN THE 21ST CENTURY

2025-08-28T06:56:08+00:00

The 21st century has mainly witnessed a huge exponential rise in technology-driven entrepreneurship, with tech startups emerging as one of the powerful catalysts of the process of innovation, job creation, and economic transformation. This study examines the fundamental drivers behind the proliferation of technology-based startups, analysing the socio-economic, technological, and policy-related factors that influence their development. Using a mixed-techniques approach, they study the international startup developments, funding patterns, and case research from leading startup ecosystems. The research unearths that improvements in digital infrastructure, full-size internet penetration, and evolving customer behaviour have improved the emergence of era startups. Additionally, the democratisation of equipment and access to international markets has empowered entrepreneurs to innovate with minimal assets. However, challenges, which include regulatory constraints, funding obstacles, and marketplace competition, continue to be vast. The study concludes with coverage and strategic suggestions to support the sustainable growth of tech startups globally.
Keywords: , ,, ,

MICROSERVICES WITH ASP .NET CORE AND OOP DESIGN PRINCIPLES

2025-08-28T06:58:09+00:00

Microservices are now the de facto architecture to build scalable and resilient systems and it is even more apparent in the age of distributed computing. ASP.NET Core is one of the top frameworks to build such systems because of the modularity, cross-platform ability, and capability to work in cloud-native settings. At the same time, object-oriented principles (OOP), particularly, object incorporation, object modularity, and polymorphism, remain strong practices of the software design, but there is a research gap to correlate them to microservices systems.
This paper investigates the implementation of OOP concepts in ASP.NET Core-based microservices and examines how this approach can be used in the area of system design, ease of maintenance, and performance amelioration. Thematic synthesis of peer-reviewed literature, case studies, and technicaldocumentation were used as a qualitative research method. Architecture patterns that are based on a code, including Dependency Injection, Domain-Driven Design, and asynchronous messaging, have been examined.
The results indicate that the great benefit of OOP and principles, implemented through SOLID-conformant service, the interface-based approach to contracts, and through the Repository or Factory design patterns can be achieved through the scalability of the application, decoupling, and long-term support. Experimental deployments with Docker and Kubernetes demonstrated up to 35% performance improvement and improved fault tolerance. Also, the flexibility of ASP.NET Core makes it possible to merge it with other more modern paradigms like declarative logic, IoT-based workflows, and an event-driven architecture.
These findings affirm that OOP is not just comprehensible with microservices in ASP.NET Core but is anyhow structurally fundamental to their effectiveness since it can offer a perfect theoretical framework or even practice useful benefits in developing long-lasting enterprise-level distributed frameworks.

Responsible AI in Real-Money Gaming: Embedding Regulatory Constraints in Personalization Algorithm

2025-08-28T07:24:33+00:00

The use of AI-powered personalization in real-money gaming has raised significant ethical and regulatory concerns in recent years. These systems are usually designed to maximize user en- gagement with the platform, which leads to higher monetization, but this sometimes comes at the cost of opaque and potentially discriminatory targeting systems, which risk exploiting user behavior, particularly among the vulnerable population. This study aims to address these issues by developing a regulatory-aligned framework that integrates ethical constraints such as fairness, transparency, and harm mitigation directly into the design of personalization algorithms. Draw- ing on secondary qualitative sources such as regulatory frameworks, peer-reviewed literature, and documented platform practices, this study analyzes current standards and gaps in industry practice. These insights inform the creation of the Responsible Personalization Framework (RPF), which operationalizes regulatory principles through concrete design patterns. The findings re- veal that widely adopted personalization strategies often contradict responsible AI principles. However, the proposed framework, featuring mechanisms like Harm-Aware Personalization and the Offer Parity Rule, offers actionable, regulation-compliant solutions for ethical personaliza- tion. Overall, the study demonstrates that embedding ethical and regulatory safeguards at the design stage, instead of post hoc reactive enforcement, enables the development of safer, fairer, and more responsible AI systems in the context of real-money gaming.

Data Storage and Speed: Why Some Businesses Use Both MongoDB and Aerospike

2025-08-28T07:27:13+00:00

This paper presents a research study that seeks to understand the performance optimization that can be established through the co-deployment of MongoDB and Aerospike databases in contemporary applications and their relative advantages concerning supporting different types of workloads. The document-based NoSQL database is MongoDB, which excels in schema flexibility, extensive document manipulation, and complex queries, making it ideal for dynamic setups like user profiles or product inventory examples. Aerospike is a low-latency, high-performance key-value store, tailored toward low-latency, high-throughput workloads, and specifically to use cases such as session management and real-time data feeds. Operating Aerospike and MongoDB concurrently enables an organization to leverage the support of both databases to query and make high-volume transactions whenever a workload comes along. The study investigates the realization of such high-performance architecture through the aggregation of the said systems and enlists the benefits and compromises of this dual-stacking technique. Some significant areas of future research are on understanding the performance of various configurations, making latency and storage costs controllable, and the predictive modelling of workload behavior using machine learning-based techniques such as correlation-based feature selection (CFS) and principal component analysis (PCA). Also, the paper points out a lack of recent empirical knowledge about dual-stack deployment. It presents future work opportunities, including the use of larger datasets, comparisons of MongoDB and Aerospike with additional NoSQL systems, and performance tuning automation via online learning techniques. Results indicate that co-deployment of MongoDB and Aerospike is optimal when there is a need to operate flexibly querying systems that must achieve reliable, low-latency behavior, resulting in a developer-friendly system that is operationally scalable.

Secure Messaging Protocols for Transactional Health Notifications

2025-08-28T07:29:08+00:00

Secure messaging in healthcare in the digital age is paramount for safeguarding sensitive patient data, maintaining patient trust, meeting regulatory standards, and delivering quality care. The increasing use of digital platforms for patient communication has made health information more important to be safeguarded. This paper describes some secure messaging protocols, including their definition, significance, key features, and types, particularly using secure messaging protocols in transactional health notifications (such as appointment reminders, and test results). In addition to discussing the challenges faced when implementing a secure messaging system, the paper also discusses the challenges of integrating it with legacy systems, privacy concerns, and others. It also clarifies the ethical and legal ramifications of secure communication in healthcare and the obligation imposed on healthcare providers to refrain from exposing patient data to unauthorized or post breaches. Finally, the paper discusses emerging technologies, including artificial intelligence and blockchain, and how these technologies, along with the changes in the regulations, will change the future of secure messaging in health care, enhancing security, facilitating processes, and supporting the compliance of the whole industry.

Edge AI for Real-Time ICU Alarm Fatigue Reduction: Federated Anomaly Detection on Wearable Streams

2025-08-28T07:34:12+00:00

ICU alarm fatigue is a serious problem in the healthcare sector since healthcare practitioners are overwhelmed by the number of alarms that are triggered, and most of them are false positives. It causes desensitization, a delay in response, and compromises patient safety. Federated learning and Edge AI present a viable solution to curbing alarm fatigue through optimization of alarm management systems and effective real-time patient monitoring. Edge AI enables data processing on wearable devices, ensuring that alerts occur promptly and with minimal delay. Federated learning allows machine learning models to be learned on decentralized and secure patient data without direct access to protected health information, maintaining privacy and customizing alarm limits. This paper discusses the possibilities of federal anomaly detection in wearable devices relating to ICU patients, in the context of real-time detection of health anomalies like abnormal heart rates and oxygen saturation. The objective is to evaluate how these technologies can streamline alarm systems by minimizing false alarms, while focusing on the important event. Important insights show the potential of Edge AI to enhance healthcare processes and deliver insights capable of driving interventions with minimal input. Federated anomaly detection is an innovative solution that can improve the work of an ICU, both in terms of operational efficiency and patient safety. This analysis seeks deeper research and the application of these technologies to clinical practice to reach their full potential.

Integrating Threat Intelligence with DevSecOps: Automating Risk Mitigation before Code Hits Production

2025-08-28T07:35:52+00:00

The combination of Threat Intelligence (TI) and DevSecOps pipelines allows organizations to automate risk reduction measures in the code before production. This paper outlines the complete picture of ingesting, normalizing, and operationalizing TI feeds, including both commercial and open-source options, as well as those based on honeypots within CI/CD pipelines. It has defined standardized data formats (STIX, TAXII) and parsers for extracting indicators of compromise (IOCs), and tactics, techniques, and procedures (TTPs). Policy-as-code gates (Open Policy Agent, HashiCorp Sentinel) allow real-time blocking during a build to occur with configurable severity. A representative selection of microservices and open-source applications was evaluated experimentally, showing 45% fewer vulnerable builds and 30% smaller mean time to remediate (MTTR) with only a modest pipeline latency overhead of 5%. Case studies provide descriptions of Kafka-based ingestion topology, enhancement through VirusTotal and AlienVault OTX, and blended dashboards with Grafana and ELK. The areas covered in the discussion are the issues in false positives, feed-quality SLA, and performance optimization via parallel processing and caching. Future research directions will be predictive blocking using AI and deep learning, auto-tuning using closed-loop feedback, multi-cloud and service-mesh integrations, and joint risk scoring with SAST/DAST tools. The results validate that an automated TI integration turns security into a scalability enabler of secure, agile software delivery. It has granular auditing trails that would help comply with GDPR and PCI DSS.

Machine Learning for Dynamic Pricing Strategies in E-Commerce and Physical Retail

2025-08-28T07:37:26+00:00

Dynamic pricing is a pricing strategy involving the change of prices based on demand, market conditions, competition, and consumer behavior. Dynamic pricing is dissimilar to static pricing, as prices do not stay constant and are adapted in real-time to market variations, thus achieving more revenue and vastly improving customer satisfaction. Machine learning (ML) is an integral ingredient for dynamic pricing as it allows algorithms to process vast amounts of data to adjust prices in real time. Instead, it allows companies to choose context and personally sensitive prices that improve their e-commerce and physical retail competitiveness. In e-commerce, for example, Amazon uses dynamic pricing algorithms that change prices daily based on market conditions and competitors' actions. At the same time, physical retail stores without Internet of Things (IoT) sensors and digital displays are rapidly converging to real-time price optimization. As with pricing systems, ML-based dynamic pricing systems are advantageous as they are automated and can respond faster to market changes. Problems like fairness, confining data privacy, and customer judgment on pricing, among other things, may still exist. Dynamic pricing models will become more transparent and precise with future trends in machine learning, like explainable AI and quantum computing. ML is also integrated with emerging technologies like augmented reality and blockchain to customize pricing strategies further. With the dynamic state of retail, machine learning will play a vital role in optimizing dynamic pricing models and improving business profitability.

"BIM-to-Field" Inspection Workflows for Zero Paper Sites

2025-08-28T07:39:06+00:00

The article presents practical ideas of BIM-to-Field solutions for zero-paper construction sites. Approaches favor model-based tagging of inspection points in IFC/COBie exports, schema Serialization as compact JSON, on-device SQLite (offline-first, delta sync, and conflict resolution), QR/RFID localization, mobile UI (voice input and photo markup), and real-time QA dashboards. Microservices and IFC viewers bridge the capture in the field communications to a shared data environment. There are two performance-validating pilots: a residential tower in 42 stories, and a rural bridge program. Teams in the tower completed 1,830 inspections in 4 weeks at an average sync time of ~4.2 seconds, detected 38 percent extra defects, and reduced unplanned rework to 2.8 per week (down 5.7 per week). On the bridge system, inspectors operated offline with <150 ms local scan and <90 s upload; safety-critical decision speed decreased 63.2% (18.4 to 6.7 hours), and non-safety close rates rose 47%. Outcomes demonstrate that the error rate decreased to 0.8 percent (7.2; =49) and boosted cycle time by 18.3 minutes (27.5 -18.3; -33). The economics are 145k capex, 3.2k/month opex, 335k annual net benefit, five-month payback, and ~460% three-year ROI. Implications: increased traceability, quicker decision-making, and less paperwork. The other issues to be overcome include schema matching, segmentation of the model applicable to resource-limited devices, device/peripheral management, and change. This offers a robust roadmap and presages AI-aided visual QA and IoT/drones and signed, append-only inspection plans, in the longer term, interoperability.

Budget Reallocation Strategies for Programmatic Advertising Using Reinforcement Learning and Historical ROAS Signals

2025-08-28T07:40:51+00:00

Programmatic advertising campaigns need budget changes because of performance and consumer actions. Usual methods with set budgets plus fixed rules do not work when audiences or competition change. This paper describes a reinforcement learning framework that reallocates budget to different times of the week and day. The framework uses historical Return on Ad Spend but also sales lift data for optimization.
The system uses a Deep Q-Network (DQN) with several hidden layers and experience replay for budget work. This study employs a DQN model, trained on e-commerce campaign data. The data held $1.5 million in advertising spend and 1.2 million impressions across 12 months. The RL agent targets segments that show high expected ROAS, plus it keeps cost limits. The agent adjusts budgets every hour - this approach raised ROAS by up to 21.8 % and dropped cost-per-acquisition (CPA) by 18.4 %. This occurred when compared to a constant baseline but also a rule-based system. A web-based Opportunity Dashboard also shows future sales growth and the best budget changes as they happen.
The results show that over a variety of time periods, RL-driven budget reallocation significantly increases campaign effectiveness and reduces unnecessary advertising expenditures. The study concludes with a detailed analysis of deployment factors, including whether to expand to multi-channel campaigns on various digital advertising platforms, how to create rewards, and how frequently to retrain the model.

DWDM Optimization: Ciena vs. ADVA for <50ms Global finances

2025-08-28T07:45:52+00:00

Optical transport with ultra-low latency has become a key factor in competition for high-frequency trading and real-time settlement, pushing the tolerable end-to-end (round-trip) delay on intercontinental routes down to less than 50 ms. This study compares the two leading open-line DWDM ecosystems: a 6500 with WaveLogic 5 Nano and the ADVA FSP 3000 TeraFlex. This system comprises five nodes of a global ring located in New York, London, Frankfurt, and Singapore. Launch power, baud rate, FEC overhead, amplifier spacing, and FlexGrid slot width are simultaneously optimized using a multi-objective genetic algorithm, which minimizes latency, maximizes OSNR, and achieves a given spectral efficiency while subject to ten-year power and capital cost constraints. GNPy models physical layer impairments, such as propagation delay and data capture, capture on 100 km recirculating fiber loops with nanosecond-grade timestamping. Ciena meets the 50 ms round-trip target and has a consistent 0.7–1.9 ms advantage on trans-oceanic spans, partly due to shorter FEC codewords and shallower DSP pipelines, as well as hybrid Raman/EDFA repeaters that reduce regeneration events. On shorter European legs, the ADVA achieves comparable performance, while it results in marginal CAPEX savings when electricity or leased spectrum is a prominent component of operating costs. Although Ciena is ranked 0.87 against ADVA, 0.83 when weighted composite fitness scores are tailored to focus on latency (35%), OSNR, protection speed, spectral efficiency, and total cost, sensitivity sweeps illustrate cases where ADVA could outperform Ciena. Hybrid Raman amplification, byte-aligned FEC, FlexGrid slot adaptation, and 1+1 protection are suggested by robust design guidelines carefully distilled from the Pareto frontier to ensure deterministic performance. They establish a reproducible benchmark and inform future adoption across the global financial backbones, featuring terabit-class engines, AI-driven power tuning, pluggable ZR+ optics, and quantum-safe encryption.

PAIR MEAN CORDIAL LABELING SOME FAMILIES OF SPECIAL GRAPHS

2025-08-28T08:54:09+00:00

In this paper we study Pair Mean Cordial (PMC) labeling is type of graph labeling in which each edge receives a label determined by the mean (average) of the integer labels assigned to its endpoints and the resulting edge labels are distributed as evenly as possible. We investigate the existence of PMC-labeling for several well-known special graphs like gear graphs, bistar graphs, book graphs, barbell graphs, and their unions. For each class, we determine necessary and sufficient conditions under which a PMC-labeling exists, and provide constructions or counterexamples accordingly. We also examine how the PMC-property behaves under union operations and identify criteria that preserve cordiality.

Decision-Making Framework for Intrusion Detection in Networks Using XGBoost-Based Feature Selection and Deep Neural Networks

2025-08-28T12:08:40+00:00

The growing complexity and frequency of cyberattacks have rendered intrusion detection a vital aspect of network security. The conventional Intrusion Detection Systems (IDS) tend to be challenged with high-dimensional data and poor capability to detect variegated patterns of attacks. To address such issues, this study introduces a decision-making framework for intrusion detection through the integration of XGBoost-based feature selection and Deep Neural Network (DNN) classification. The suggested methodology was tested on two popular benchmark datasets: NSL-KDD and CIC-IDS 2017. Feature selection was first carried out by using the XGBoost algorithm to eliminate redundant and less informative features while preserving the most relevant attributes. This step enhanced the efficiency and accuracy of the model. The features were then classified using a DNN, which took advantage of its robust representation learning ability in identifying different types of attacks. Experimental results confirm the efficacy of the presented XGBoost–DNN approach. On the NSL-KDD dataset, the model performed with an accuracy of 99.69%, successfully identifying major attack categories such as. Likewise, on the CIC-IDS 2017 dataset, the model performed at 99.25% accuracy, showing excellent accuracy and recall for several contemporary attack types. These comparison of results with previous work demonstrate the strengths of the proposed methodology in dealing with heterogeneous and complex network traffic data.

Emerging Trends in Sentiment Analysis within Natural Language Processing for Social Media

2025-08-29T02:28:44+00:00

Sentiment analysis, a vital domain of natural language processing (NLP), aims to classify textual data into sentiments such as positive, negative, or neutral. With the increased use of social media for expressing opinions, efficiently analyzing these sentiments has become crucial for organizations in sectors ranging from marketing and politics to finance and customer service. This paper reviews current methods and advancements in sentiment analysis, highlighting traditional and modern approaches, challenges with social media data, and proposing a robust, scalable framework using deep learning and transformer-based models. Comprehensive experimentation validates the effectiveness and outlines the future scope for more nuanced and multilingual sentiment evaluation.

MALICIOUS APPLICATION DETECTION USING STACKED ACLR

2025-08-29T02:36:43+00:00

With the increasing prevalence of mobile applications, security threats posed by malicious apps have escalated, leading to data breaches and privacy violations. Traditional signature-based detection methods struggle against evolving attack techniques. This project proposes an advanced Malicious Application Detection System using a stacked ACLR deep learning model (Artificial Neural Networks, Convolutional Neural Networks, Long Short-Term Memory, and Recurrent Neural Networks) to enhance detection accuracy. The system leverages deep learning for robust feature extraction from permissions, API calls, network activity, and metadata, providing real-time and adaptive classification of applications. By integrating CNNs for spatial pattern recognition, LSTMs for sequential analysis, and RNNs for behavioral modeling, the proposed system significantly outperforms conventional machine learning methods. The experimental results demonstrate higher accuracy, reduced false positives, and improved detection of zero-day attacks. This project contributes to an intelligent, scalable, and automated cybersecurity framework, strengthening protection against malicious mobile applications in dynamic digital environments..

SKIN CANCER DETECTION USING DEEP LEARNINIG

2025-08-29T02:38:48+00:00

Skin cancer is one of the most common and life-threatening forms of cancer worldwide, with its incidence increasing due to factors such as ozone depletion and prolonged ultraviolet (UV) exposure. Early detection plays a critical role in improving survival rates; however, conventional diagnostic methods, such as dermatological examinations and biopsies, are time-consuming, expensive, and prone to subjectivity. Recent advancements in artificial intelligence, particularly deep learning, have shown significant potential in addressing these challenges by enabling automated, accurate, and efficient skin cancer detection.This research proposes a deep learning–based skin cancer detection system using Convolutional Neural Networks (CNNs) for the classification of dermoscopic images. The system leverages the HAM10000 dataset to classify seven skin cancer types, including melanoma and non-melanoma variants. Through preprocessing techniques such as resizing, normalization, and data augmentation, the proposed model ensures improved robustness and generalization. Furthermore, advanced architectures like ResNet50 are explored to enhance classification accuracy.The experimental results demonstrate that CNN-based approaches achieve high precision and reliability, significantly outperforming traditional methods. By integrating such models into mobile and cloud-based healthcare platforms, the system offers scalability, accessibility, and real-time diagnostic support, making it particularly useful in remote and resource-limited regions. The proposed work highlights the potential of deep learning to revolutionize dermatological screening, improve diagnostic accuracy, reduce reliance on invasive procedures, and ultimately contribute to better patient outcomes and global healthcare standards.

Transformative Randomized Decision Trees for Sleep Disorder Classification from Health Life-Style Data

2025-08-29T13:36:27+00:00

Sleep disorders, such as insomnia and sleep apnea, impact a substantial portion of the global population, with insomnia affecting approximately 10% of adults and sleep apnea influencing up to 3%. Despite these significant statistics, existing diagnostic approaches face challenges such as incomplete datasets and suboptimal classification accuracy. Traditional methods often struggle with differentiating between various sleep disorders due to limitations in feature extraction and classification techniques. To address these challenges, a novel Sleep Disorders Classification (SDC) framework is proposed. This framework incorporates advanced preprocessing techniques to enhance data quality and uses Incremental Eigen Values Analysis (IEVA) for efficient feature extraction, which dynamically reduces the dimensionality of the dataset. The framework also employs Randomized Decision Trees (RDT) classification to accurately distinguish between insomnia, sleep apnea, and normal sleep conditions. By integrating these advanced methods, the SDC framework aims to improve the accuracy and reliability of sleep disorder diagnoses.

Machine Learning-Powered Dynamic Analysis Framework for Effective Android Malware Detection

2025-08-29T13:40:10+00:00

Since Android smartphones are so widely used, hackers have turned them into their main target, which has resulted in a sharp increase in sophisticated mobile malware. Only static analysis techniques are not enough to detect evolving threats due to obfuscation and code transformation techniques. We explore a dynamic analysis-based approach for Android malware detection using machine learning algorithms in this research, leveraging the CICMalDroid 2020 dataset. Various classifiers, including Random Forest, Decision Tree, Naive Bayes, Logistic Regression, AdaBoost, Extra Trees, and Gradient Boosting Machine, were trained using dynamic features that captured runtime behavior, such as system calls, API calls, and other runtime behaviors. An extensive assessment of model performance is made possible by the dataset's inclusion of a wide range of malware families and benign applications. Common metrics like Accuracy, Precision, Recall, F1-score, Specificity, Matthew Correlation Coefficient, Cohen Kappa, and ROC Score were used to evaluate the models. The results of experiments show that machine learning models trained on dynamic features are capable of accurately differentiating between malicious and benign applications, with the highest detection accuracy being attained by the Random Forest classifier.

Smart Detection Machine Learning for Affordable Chronic Kidney Disease Screening

2025-08-30T04:27:27+00:00

Chronic kidney disease (CKD) is a major global health issue, one that is increasingly prevalent in both advanced and emerging economies. Especially in under-resourced settings, the lack of specialized tools greatly complicates early diagnosis. This study proposes some methods for machine learning (ML) to determine key features of CKD that could be integrated into low-cost diagnostic screening tools accessible to primary level healthcare providers We use and test various machine learning approaches such as Random Forest and Support Vector Machine (SVM) and a custom ensemble model to analyze a large set of clinical factors for the disease. In this case, the provided framework achieves CKD prediction with 97.8% accuracy, 96.5% sensitivity, and 98.4% specificity from just the most basic clinical parameters available — outcomes that are achievable by any standard. Our research illustrates that diagnostic tools based on ML can accurately assess vulnerabilities for CKD with Fundamental clinical indicators can improve patient care and speed up response time in resource-constrained environments. Effective implementation provides a proof- of-concept that balances diagnostic precision with reasonable access, showing it is adaptable within existing healthcare systems.

Lightweight Network Intrusion Detection in IoT Environments Using Machine Learning

2025-08-30T04:38:56+00:00

With the rapid expansion of IoT devices, network intrusion security has indeed become a major issue. This work introduces a machine learning-based IDS using the IoTID20 dataset representing the imbalanced real-world network traffic. Data preprocessing pipeline converts categorical attributes into numerical attributes through Label Encoding, normalizes data via Min-Max Scaling, and selects features using Pearson Correlation. Three classifiers, the Support Vector Machine (SVM) based on One-Versus-Rest, Decision Tree, and Random Forest, have been evaluated. The SVM achieved an accuracy of 83%, while Decision Tree and Random Forest achieved 89.06% and 96.02% respectively. It is evident from the above results that with the correct preprocessing and feature selection, the detection performance can be greatly affected even on imbalanced datasets. The built system thus functions as an accurate yet lightweight IDS solution that can be deployed in resource-limited IoT environments.

Sentiment-Enhanced Trading Deep Q-Network: Advancing Financial Trading with Deep Reinforcement Learning

2025-08-30T04:58:58+00:00

This paper presents the Sentiment-Enhanced Trading Deep Q-Network (SETDQN), a novel deep reinforcement learning (DRL) framework for optimizing financial trading strategies. By integrating historical price data, technical indicators, sentiment embeddings from social media platforms, and macroeconomic indicators, the SETDQN maximizes the Calmar ratio, a risk-adjusted performance metric. Trained onS&P 500 ETF (SPY) data from 2010–2020 and tested on 2021–2024, the SETDQN achieves a 17.5% annualized return and a 2.1 Calmar ratio, surpassing traditional strategies like recurrent reinforcement learning (RRL), technical analysis, and buy-andhold. The implementation, provided in Python, is reproducible on Kaggle, incorporating realistic market frictions such as transaction costs and bid-ask spreads. This work advances DRL applications in finance, offering a scalable and robust framework for algorithmic trading.

A COMPREHENSIVE STUDY OF CNN-BASED FACIAL EMOTION RECOGNITION IN IMAGES AND VIDEOS

2025-08-30T05:50:26+00:00

Human emotional states are categorized using facial emotion recognition. Sorting each face image into the seven classes of facial emotions is the goal. Convolutional Neural Networks (CNNs) are employed in the emotion classification process. Real-time videos and A range of grayscale images from the dataset are captured for input. Next, the CNN's sequence of convolution and purpose of pooling layers is information abstraction, and the SoftMax layer is used for classification. A few methods are used to address the model's overfitting issue, including dropout, cluster standardization, and L2 regularization. The facial expression dataset from the Picture Folders (fer2013) collection is applied to experiments, and our model performs more accurately in predicting individual emotions than previous research has. Furthermore, the model exhibits good performance in predicting the mood of every picture in the live video stream.

AN APPLLICATION OF MULTIPLE RENEWABLE ENERGY SYSTEMS WITH PLURAL GOAL OPTIMIZATION

2025-08-30T08:00:43+00:00

The Proposed System is a mixed system with features of renewable nature and hybrid nature. It consists of panels which are photovoltaic in nature (photo voltaic cells), turbines which operates with wind (wind turbines), and also it will have non- stop production of hydrogen through green energy route using reformers. Genetic algorithm is applied for optimization and is validated. In the hypothecated system the storage of hydrogen the annualized cost for storing the hydrogen is minimized and the energy wastage is reduced by 65%. A detailed sensitivity analysis conducted on the effect of cost variation, impact of different kinds of storage, and various capacities on system performance. the issues related to green criteria as environment sustainability has become the need of the hour optimization of cost, and green criteria are taken as objectives in the paper Multi-objective formulation is done, and a suitable metaheuristic is applied to solve the problem.

SECURE AND OPTIMIZED CLOUD AUTHENTICATION USING XMSS WITH HTM-BASED HYBRID CRYPTOGRAPHIC KEY MANAGEMENT

2025-09-01T05:39:53+00:00

With the rapid adoption of cloud computing services, ensuring secure, efficient, and scalable authentication mechanisms has become increasingly critical. Traditional cryptographic techniques are often vulnerable to emerging threats, especially quantum attacks and side-channel vulnerabilities. To address these challenges, post-quantum cryptographic algorithms and hardware-based optimizations are gaining momentum. Conventional authentication methods in cloud environments often fail to ensure forward security, resistance to quantum-level threats, and protection against side-channel attacks. The lack of isolation during cryptographic execution further exposes private keys and signature schemes to timing and leakage attacks. This study introduces an enhanced cloud authentication model that integrates the eXtended Merkle Signature Scheme (XMSS), a hash-based, forward-secure, and post-quantum resistant signature scheme, with Hardware Transactional Memory (HTM). XMSS guarantees quantum-resistant authentication, while HTM ensures secure execution by isolating sensitive cryptographic operations in protected hardware regions. The system further employs hybrid cryptographic key algorithms to optimize the processes of key generation, signing, verification, and key storage. The HTM engine assists in minimizing leakage and enhancing efficiency, enabling secure communication between cloud users and service providers. Simulation results show a substantial improvement in key verification speed, signature integrity, and resistance to quantum and side-channel attacks. Compared to existing schemes such as RSA, ECC, and SPHINCS+, the proposed method offers a 30–40% performance boost in security-sensitive operations, with a reduced risk of cryptographic key exposure and computational overhead.

Attention-Based Cross-Modal CNN Using Non-Disassembled Files for Malware Classification

2025-09-01T06:01:42+00:00

Battling the spread of perilous programming renditions relies first upon the order of malware. This paper handles this issue by recommending a new technique utilizing a “Convolutional Neural Network (CNN)”- based model to group malware occasions into families without relying upon dismantled code, thus inclined to botches. Rather, the model purposes non-dismantled paired documents coordinating two modalities: primary entropies and malware pictures. These modalities offer a few points on the information, subsequently further developing order exactness. Highlights from the two modalities are productively coordinated utilizing a cross-modular consideration process, hence diminishing their different limitations. With extraordinary accuracy of 98%, the recommended model is contrasted and customary strategies including "VGG16, CNN, and XGBoost". Alongside embracing the Xception model, which maybe surpasses close to 99% accuracy, troupe methods including Casting a Voting Classifier and Decision Tree are explored to further develop execution more. An easy to understand frontend produced for testing and verification utilizes likewise a Flask system. This sweeping technique increments client availability and security in malware examination notwithstanding malware order accuracy.

Evaluating Intelligent Methods for Software Risk Prediction: An Empirical Analysis

2025-09-01T07:00:08+00:00

Software development involves significant uncertainty due to unexpected events occurring at various stages of the software development lifecycle. As software size and complexity increase, the risk of project failures also rises. These unexpected events, known as software risks, stem from various factors throughout the development process. Effective risk management during the early phases is crucial to ensure a high-quality final product. Traditional risk assessments rely on human expertise and past experience, which can be subjective and less reliable. This study employs machine learning methods to predict software risks using historical data, aiming for early and accurate risk detection. Five machine learning models were evaluated alongside multiple feature selection techniques to improve prediction accuracy. Experiments were conducted using publicly available software risk datasets. Results show Support Vector Machine (SVM) model achieved highest classification accuracy of around 80%. Among feature selection methods, Mutual Information demonstrated superior performance across evaluated models, enhancing effectiveness of risk prediction.

Deep Learning and Optimization-Based Pest Detection in Peanut Crops Using CNN, MFO, and EViTA

2025-09-01T12:18:22+00:00

The rapid advancement of Vision Transformer (ViT) methods has proven highly effective in image classification and identification tasks. This paper introduces an Enhanced Vision Transformer Architecture (EViTA) tailored specifically for pest identification, segmentation, and classification. Building upon ViT's strengths over Convolutional Neural Networks (CNNs), EViTA aims to improve accuracy in pest image prediction. The methodology incorporates preprocessing techniques such as Moth Flame Optimization (MFO) for image flattening and normalization, along with a dual-layer transformer encoder for integrating pest image segments of varying sizes. Extensive experiments using three pest datasets affecting peanut crops demonstrate the efficacy of EViTA, achieving promising results. Furthermore, the exploration of additional techniques such as DenseNet, InceptionV3, and Xception TL models suggests potential accuracy improvements beyond 94%. Additionally, the integration of Flask framework enables the development of a user-friendly front end for testing with authentication. EViTA presents a novel approach to pest identification with significant implications for enhancing pest management and agricultural practices. Further research and refinement hold promise for advancing EViTA's capabilities in pest identification tasks.

A Secure and Intelligent Healthcare Chatbot Using Blockchain And AI

2025-09-02T05:21:30+00:00

In recent years, the healthcare industry has increasingly adopted digital solutions to improve efficiency, accessibility, and patient experience. However, the rapid digitization of health information raises critical concerns around data privacy, security, and unautho- rized access. This project introduces a comprehensive web-based system that securely manages and stores electronic health records using Ethereum blockchain technology. Developed using Python Django, HTML, CSS, JavaScript, Bootstrap, and MySQL, the system ensures that all medical data is handled in a secure, transparent, and decentralized manner. The core of this system is built upon blockchain technology powered by Ethereum and tested on the Ganache local blockchain environment, which enables tamper-proof storage and retrieval of sensitive patient information through smart contracts.
The application comprises three main user portals: Admin, Doctor, and Patient. The Admin can register, manage, and monitor both doctors and patients while reviewing user feedback to ensure the smooth operation of the platform. Doctors can manage their profiles, handle appointment scheduling, and communicate di- rectly with patients through a built-in messaging system. Patients can explore doctor profiles, book appointments, view their medical history, and interact with doctors in real-time. One of the key highlights of this system is the integration of a healthcare chatbot using the Perplexity API, which assists patients by responding to common medical queries, thus improving accessibility and reducing dependence on direct consultations for preliminary concerns.
Blockchain plays a vital role by providing a secure environment for health record storage. Once a record is uploaded, it is hashed and stored immutably on the Ethereum blockchain using smart contracts, ensuring that the data cannot be altered or deleted by any unau- thorized party. This approach significantly strengthens the trust, integrity, and confidentiality of patient records. By combining modern web technologies, artificial intel- ligence, and blockchain, this project demonstrates an efficient, user-friendly, and highly secure solution for managing digital health records in a scalable and future- ready manner.

Investigating The Use of Higher-Order Spectral Features for Graph Analysis and Machine Learning Tasks

2025-09-02T11:35:41+00:00

Spectral approaches have been a key part of graph research for a long time, however most Graph Convolutional Networks (GCNs) use constrained Laplacian eigenvalue representations that don't provide higher-order structure information. This research looks at application of higher-order spectral features to improve machine learning tasks that use graphs. Based on the basic ideas of spectral graph theory. This research presents a spectral-feature-augmented GCN system that incorporates spectral entropy, wavelet signatures, heat kernel embeddings, and eigenvalue moment features to improve representation learning at both node and graph levels. This research looks at how well higher-order spectral characteristics may help with graph analysis and learning tasks. This system captures both local and global topological information by adding spectral features as spectral moments, entropy, wavelets, and heat kernel signatures to standard graph neural network (GNN) designs. The work uses the spectral graph theory to provide theoretical reasons and tests the actual advantages on benchmark datasets including Cora, MUTAG, and CiteSeer. The suggested spectral-feature-augmented GCN performs better than regular GCNs in node classification, graph classification, and link prediction, with significant gains in accuracy. These results show that there is a trade-off between the cost of computing and the performance of learning. They also show how important deep spectral information is in real-world graph learning pipelines.

Invisible Gaze in the Algorithmic Age: A Mixed-method Exploration of Legal Boundaries and Social Justice in Non-consensual Surveillance in China

2025-09-02T13:44:24+00:00

This study explores non-consensual surveillance in China, focusing on the intersection of digital privacy, algorithmic bias, and social justice. This qualitative study uses a mixed-method approach to examine participants' awareness of digital surveillance, their experience with data collection, and their perception of how it affects marginalized groups. The obtained results indicate that there is high awareness of surveillance on social media and e-commerce sites, in which concerns of algorithm discrimination and unequal targeting of women, LGBTQ+, and racial minorities are a significant concern. Although China has a law addressing the Personal Information Protection Law (PIPL), the paper finds an information gap on legal awareness and enforcement. The study recommends intensifying regulatory activity, greater openness in collecting data, and additional public awareness of digital privacy rights. The research highlights the importance of a comprehensive approach to digital surveillance that safeguards individual and social justice rights.

Comparative study of hybrid cryptographic solutions (classical + post-quantum) for maintaining performance and security in IaaS in the face of quantum threats in a transitional period towards fully PQC1 solutions

2025-09-02T15:12:33+00:00

It is imminent that the arrival of quantum computing is a real threat to data security in IaaS, in transit as well as at rest. To find cryptographic schemes capable of dealing with powerful quantum algorithms such as Shor and Grover, this article proposes an evaluative study of hybrid cryptographic solutions, which combines classical algorithms such as RSA2, ECC3, with post-quantum algorithms such as Kyber and Dilithium, capable of maintaining the same performance levels in an IaaS environment, and at the same time face the power of quantum systems. To do this, several metrics will be analyzed, including latency, CPU4 load, RAM5 saturation, and resistance to quantum attacks to have an objective idea of their possible adaptation (hybrid solutions) in the different IaaS environments of our main IaaS resource providers around the world, with a view to ensuring a transition to fully PQC solutions.

INFORMATION-SEEKING BEHAVIOR AS A FEATURE FOR MACHINE LEARNING–DRIVEN USER CLASSIFICATION

2025-09-03T07:30:31+00:00

The rapid expansion of digital platforms has led to diverse patterns of online user behavior, particularly in how individuals search for, access, and interact with information. Understanding these patterns is essential for applications such as personalized recommendations, targeted marketing, cybersecurity, and digital forensics. This paper presents a machine learning approach to the classification of online users by exploiting their information-seeking behavior. By analyzing search queries, browsing histories, and navigation trails, the proposed framework identifies latent behavioral features that serve as strong predictors of user categories. Advanced algorithms such as decision trees, support vector machines, and neural networks are employed to capture both linear and nonlinear patterns in user behavior. Feature engineering techniques, including term frequency–inverse document frequency (TF-IDF) representations and sequential activity modeling, are used to enhance classification accuracy. Experimental results demonstrate that integrating behavioral features with machine learning significantly improves classification performance compared to conventional demographic-based approaches. This research highlights the potential of behavioral data as a valuable resource for accurate and adaptive online user classification, with implications for recommender systems, online security, and human–computer interaction.

FUSING TEXT, IMAGES, AND USER METADATA: A HYBRID APPROACH TO FAKE NEWS DETECTION ON TWITTER

2025-09-03T07:31:52+00:00

Such extensive spread of misinformation via social media is a significant threat to trustworthiness, the health of the citizens, and healthy democratic institutions. Conventional methods of fake news detection are more or less limited to textual analysis and many of the time the multimodal context abundant on the web is neglected. In this research thesis, a multi-modal fake news detection framework which synergistically incorporates natural language processing and visual processing based on computer vision segments, along with user metadata analytic will be proposed to improve the accuracy of fake news detection on Twitter. The proposed system makes use of a DistilBERT model-based text analysis of the tweet content, ResNet 18-based convolutional neural network based on attached images, and feature-based knowledge of credibility based on account verification status, the number of followers, and account age. The products of these separate elements are combined into final binary classification into REAL or FAKE boundaries with the help of a Random Forest meta-classifier. The whole pipeline is implemented as a Flask web application, which allows in real-time analyzing tweet URLs. It is seen in experimental tests on both synthetic and real world datasets that the system shows strong returns in conditions of different inputs, such as when one or both modality do not work (e.g. a tweet only contains text or only an image). The suggested method provides an below-high-fidelity, socially scalable, and trusted measure of addressing the effects of misinformation between multimodal community media environments.

Study to determine the efficiency of thermal management systems in batteries

2025-09-05T05:45:30+00:00

This research focuses on the development and evaluation of the efficiency of an Intelligent Battery Thermal Management System (iBTMS) for electric vehicles in the 21st century. The study integrates multiple advanced technologies, including artificial intelligence, digital twin modeling, quantum optimization algorithms, and intelligent sensor systems. The main objective is to address thermal management issues, which remain critical limitations in enhancing electric vehicle performance and extending battery lifespan. An experimental research methodology combined with computer simulation programs was employed to assess the performance of the iBTMS compared to traditional systems under various operating conditions.
The research findings indicate a significant superiority of the iBTMS. It was able to improve heat dissipation efficiency by 24% (from 78% to 97%), reduce system energy consumption by 38% (from 450W to 280W), and extend battery life by 42% (from 2,500 cycles to 3,550 cycles) compared to conventional systems. Stability tests under different ambient temperatures (-10°C to 50°C) showed that the iBTMS consistently maintained an efficiency range of 95-98%, while traditional systems exhibited fluctuating efficiency between 35-75%. Furthermore, the digital twin computer model developed in this study demonstrated high accuracy in predicting battery temperature, with an average error of only ±0.3°C, which is considered appropriate for real-time control applications.

A Lightweight Method for Cervical Cancer Classification Using Preprocessing Pipeline and Attention-Guided Shallow-CNN

2025-09-05T05:53:17+00:00

Cervical cancer continues to be one of the most common cancers in women globally, and early detection by means of automated image classification can lead to significant improvement in survival rates. This paper introduces a novel and effective deep learning-based technique for automatic cervical cell image classification into six diagnostic classes. The model starts with a strong image preprocessing pipeline, including Contrast-Limited Adaptive Histogram Equalization (CLAHE) to enhance the visibility of lesion areas, then uniform resizing and z-score normalization to normalize image input features. These preprocessing methods ensure the input of more enhanced and normalized image data to the model, reducing the impact of contrast and scale variation. Given that the classes in the dataset are unbalanced, targeted data augmentation is used for balancing classes before training. The transformation techniques of rotation, flipping, zooming, and brightness adjustments are applied to generate synthetic samples for classes that are underrepresented, thereby reducing overfitting and improving the generalization of the model.
A shallow Compact Convolutional Neural Network (CNN) is engineered to learn discriminative features without a high computational cost as required for implementation in low-resource settings. For added classification accuracy, a new Attention-Guided Feature Modulation (AGFM) block is added to the network. This process learns to modulate spatial and channel-wise salient features and inhibit unnecessary background noise, thereby channelling the model's attention towards dysplastic areas in cervical images. Experimental results show that the designed framework is highly accurate and generalizable for multi-class cervical cancer classification. Its light-weight nature, coupled with attention-guided improvements, positions it as an ideal candidate for real-time diagnostic assistant systems, particularly in low-resource healthcare environments.

QUANTUM QUEUEING THEORY: MODELING AND ANALYSIS OF QUANTUM COMPUTING-BASED SERVICE SYSTEMS: A RESEARCH ROADMAP

2025-09-05T11:23:49+00:00

The convergence of quantum computing and classical service system modeling presents a transformative opportunity. This paper proposes the formalization of Quantum Queueing Theory (QQT), a novel framework for modeling, analyzing, and optimizing service systems where quantum processors act as servers, quantum algorithms constitute service tasks, and quantum communication channels facilitate arrivals and departures. We outline the fundamental challenges posed by quantum mechanics (superposition, entanglement, measurement, decoherence) to classical queueing paradigms. Key research directions include defining quantum analogues of arrival processes, service disciplines, and performance metrics (quantum fidelity, task success probability, decoherence-limited waiting time). We explore modeling approaches leveraging quantum stochastic processes, quantum walks, and modified Lindblad master equations. The paper details critical areas for future research: hybrid quantum-classical queueing networks, resource allocation under decoherence, stability analysis in the quantum regime, and the development of quantum-aware scheduling policies. QQT is poised to become essential for designing efficient and scalable quantum computing data centers, quantum cloud services, and integrated quantum-classical computing infrastructures.

Energy-Efficient Task Offloading and Resource Allocation for Delay-Constrained Edge-Cloud Computing Networks

2025-09-06T06:58:11+00:00

The project aims to enhance task offloading in mobile edge computing by addressing the challenges posed by limited computational resources, focusing on maximizing the number of served mobile devices (MDs) while minimizing energy consumption. To achieve this, two primary optimization problems are formulated: a quantity-driven problem to increase the number of served MDs and an energy-driven problem to minimize energy consumption, both of which are NP-hard mixed-integer nonlinear programming challenges. As a solution, a binary tree-based task offloading (BTTO) scheme is proposed, utilizing convex optimization to efficiently derive optimal task offloading decisions. The algorithm is implemented in a simulation environment, where results demonstrate its effectiveness in serving more devices with reduced energy consumption compared to existing techniques. Additionally, the project incorporates extensions such as data compression to minimize transmission time and energy use, along with security measures using SHA256 hash codes to ensure data integrity during offloading.

AI-Powered Smart Water Distribution System: An Intelligent Approach for Resource Optimization

2025-09-06T13:27:09+00:00

Water distribution systems face critical challenges including leakage detection, demand prediction, and optimization of resource allocation. This research presents a novel AI-powered smart water distribution system that integrates machine learning algorithms, IoT sensors, and cloud computing to revolutionize water management. The proposed system employs a hybrid approach combining deep learning for demand forecasting, reinforcement learning for valve control optimization, and anomaly detection algorithms for leak identification. Experimental results demonstrate significant improvements in water conservation (28%), operational cost reduction (32%), and leak detection accuracy (94.7%) compared to conventional systems. The framework's scalable architecture allows for seamless implementation across various urban water infrastructures, offering a sustainable solution to global water management challenges.

Existence and Uniqueness of Mild Solutions of Nonlinear Fractional Sum-Difference Equation with Nonlocal Condition

2025-09-08T04:47:29+00:00

In this study, we explore the solvability of nonlinear difference equations subject to nonlocal conditions. Our focus lies on identifying conditions under which mild solutions exist and are unique. To achieve this, we employ a combination of the Leray–Schauder alternative and Bihari’s integral inequalities, which together provide a robust framework for addressing nonlinearities and nonlocal constraints in discrete settings.
Beyond existence and uniqueness, the work also examines qualitative features of the solutions, including boundedness and sensitivity to changes in the initial data. These results are not only of theoretical interest but also relevant for discrete models in applied mathematics, where stability and parameter dependence play a critical role. Illustrative examples are presented to confirm the applicability and effectiveness of the theoretical developments.

GREENHOUSE YIELD PREDICTION USING FEATURE SELECTION AND ENHANCED ARTIFICIAL NEURAL NETWORK ALGORITHM

2025-09-08T04:49:20+00:00

Greenhouse yield prediction is a crucial aspect of modern agriculture, aiming to optimize production and resource management in controlled environments. Accurate yield prediction in greenhouse agriculture is paramount for optimizing resource allocation and maximizing productivity, yet the complex interplay of numerous environmental and operational factors poses a significant challenge. In this work, Improved Chicken Swarm Optimization (ICSO) and Enhanced Artificial Neural Network (EANN) algorithm is proposed. The main steps of this research includes pre-processing, feature selections, and classifications for greenhouse yield prediction. Initially min-max normalization algorithm is proposed to improve the quality of the given dataset. Then, ICSO algorithm is introduced for feature selection which selects more relevant and significant features from the given dataset. These ICSO variants, by leveraging their enhanced search capabilities and adaptability, provide robust and effective solutions for optimizing greenhouse management, resource allocation, and yield prediction. It generates best fitness values based on the higher accurate features. Finally, EANN is applied to perform the yield prediction in the given greenhouse dataset. EANN is focused to effectively manage the complexities of greenhouse environments and optimize crop yields via hidden neurons. Also, it is used for optimizing environmental controls, detecting diseases, and managing resources efficiently, ultimately leading to increased productivity and sustainable greenhouse operations. The training and testing process is conducted and it is used to provide more accurate results. The experimental results prove that the proposed APSO-EANN algorithm provides better greenhouse yield results in terms of higher accuracy, precision and RMSE values than the existing algorithms

PERCUTANEOUS INTERVENTIONS FOR ACUTE ISCHEMIC HEART DISEASE WITH CORONARY BIFURCATION LESIONS

2025-09-08T07:43:18+00:00

Bifurcation lesions in coronary arteries present a significant challenge in percutaneous coronary intervention (PCI), particularly in the setting of acute coronary syndromes (ACS). These lesions are associated with increased procedural complexity and higher rates of adverse events. This review aims to summarize current evidence regarding the optimal strategies for PCI in ACS patients with bifurcation lesions. We discuss various stenting techniques, including provisional side branch stenting, culotte stenting, crush stenting, and tap stenting, weighing the benefits and risks of each approach based on available clinical trial data. We also address the importance of intravascular imaging (IVUS or OCT) in guiding PCI and optimizing stent deployment. Finally, we highlight areas where further research is needed to improve outcomes for this challenging patient population.

LAPLACIAN MINIMUM COVERING EXTENDED ENERGY OF A GRAPH

2025-09-08T11:20:58+00:00

In the current research, we present the idea of the Laplacian minimum covering extended e´nergy of a graph EC(G) and calculate the Laplacian minimum covering extended energies of a complete graph, star graph, crown graph, cocktail party graph, and full bipartite graph.

DIABETIC RETINOPATHY DETECTION BY MEANS OF DEEP LEARNING

2025-09-09T05:14:07+00:00

Diabetic Retinopathy (DR) is a leading cause of blindness among diabetic patients. DR occurs due to prolonged high blood sugar levels that damage the retinal blood vessels. Early detection of DR is vital for timely intervention and preventing vision loss. This paper proposes an automated DR detection and classification model using deep learning techniques, specifically the InceptionV3 architecture. A publicly available dataset containing over 35,000 retinal images is used for training and validation. The model achieves a test accuracy of 97.3% and is capable of identifying five stages of DR: No DR, Mild, Moderate, Severe, and Proliferative DR. A Flask-based web interface facilitates the user-friendly deployment of the model for clinical use.

Enhancing Vehicle Routing Solutions: Custom Heuristic and Local Search Algorithms for Capacitated Vehicle Routing Problems

2025-09-09T09:12:18+00:00

When compared to traditional heuristics algorithms, local search algorithms offer significant computational advantages to solve combinatorial optimisation problems, such as the Vehicle Routing Problem (VRP). However, the actual usefulness of existing searching-based algorithms is generally limited by their inability to handle a fixed number of vehicles. These algorithms often avoid the challenging issue of assigning customers to a fixed number of available vehicles. However, in practical situations, logistic service providers need to find solutions that work with a certain fleet size and can quickly adjust to sudden changes in the number of vehicles or solve the Capacitated Vehicle Routing Problem (CVRP). Therefore, the main goal of this paper is to apply custom heuristic algorithms to solve the CVRP to determine the best vehicle delivery routes while keeping in mind each vehicle’s maximum carrying capacity. The problem is initially solved by the heuristics presented in this paper, and subsequently, specific local search algorithms are used to further improve the heuristics’ findings. The implementation of two heuristic algorithms includes one that is based on the nearest-neighbour strategy and another that is motivated by clustering nearby customers. Two custom local search algorithms are also introduced. The proposed algorithms are compared to well-used benchmarks that have undergone extensive testing and validation.

Unifying Multiple ERP Systems: A Case Study on Data Migration and Integration

2025-09-09T14:49:42+00:00

This white paper will look at the merging of three mature or older ERP environments, including SAP ECC 6.0, Oracle E-Business Suite 12.2, and Microsoft Dynamics AX 2012, into an analytics-ready and governed backbone. The program established a canonical data model (Party, Location, Item, GLAccount), and created layered landing, raw, and curated areas in Snowflake using ELT via dbt and orchestration by Airflow. Kafka consumed Change data capture to provide an ongoing harmonization and idempotency upsets. Deterministic, code-set translations and code-set translations and supervised entity resolution (blocking keys, phonetic encodings, n-gram similarities and geospatial proximity)heading Data contracts, executable tests (dbt/Great Expectations), and reconciliation packs tested completeness, conformity, referential integrity, and financial invariance across environments, and RBAC/ABAC, PII masking, immutable run manifests, and segmentation of duties accounted for SOX/GDPR controls. Practiced cutover runbooks had given results of a 7.6-hour switch of production in an eight-hour service level agreement, and no severe cases. Quality after post-maturation was significantly better: mandatory-field completeness was 99.6%, field-level conformity 98.7%, and orphan rates 0.2. DC latency achieved P95 parameter goals. The PR-AUC and F1 exceeded those of a rules-plus-logistic baseline, as gradient-boosted trees and a Siamese architecture allowed matching at millisecond latency. Business outcomes were a 1.8-day time savings on month-end close, 14.6% consolidated indirect spend, 9.2% fewer stockouts, 2.7-point invoice-accuracy improvements, 6.4% faster order creation, 41% faster analytics time-to-insight, and a decrease in the deployment failure rate (7.5% to 2.1%).

Real-Time Fraud ML on Spark Structured Streaming: Micro-Batch vs. Continuous Processing

2025-09-09T14:51:06+00:00

Digital financial transactions have become highly vulnerable to fraudulent activities due to their growing volume and velocity, which necessitates real-time fraud detection systems. This article critically examines the analysis of machine learning-based fraud detection using Apache Spark Structured Streaming, which has two processing modes: Micro-Batch and Continuous Processing. A usable pipeline is trained and constructed on the IEEE-CIS Fraud Detection dataset, which incorporates feature engineering, supervised learning models, and stream processing to perform near-real-time fraud classification. Complex transformations and fault tolerance, supported by the Micro-Batch mode, have demonstrated reliability and analytical power with a marginal increase in latency. Continuous Processing mode offers significantly improved latency and throughput, is best suited for quickly issuing alerts in high-risk environments, and is limited in its transformations and recovery actions. Comprehensive experimentation compares the two modes in terms of latency, precision, recall, resource utilization, and reliability, assessing the operations that can be achieved. The results indicated that no one mode is inherently optimal; instead, they should be chosen according to particular use case guidelines. The paper concludes with practical advice, outlining the existing inadequacies of Spark used in Continuous mode, and presents opportunities for future exploration, including active learning, graph ML, and hybrid systems that would offer the best of both worlds in real-time fraud mitigation.

The Future of DAX: Integrating Python ML Models into Power BI

2025-09-09T14:52:48+00:00

Despite rapid advancements in business intelligence (BI) platforms and machine learning (ML), the integration of DAX (Data Analysis Expressions), Python-based ML models, and Power BI remains fragmented, both academically and industrially. Current implementations that involve machine learning in Power BI are often restricted to either isolated Power Query steps or static Python visuals, which do not support real-time interactivity or native integration within the data model layer. Moreover, the absence of a dedicated function within the DAX language to invoke machine learning predictions constrains the analytical depth and agility of BI solutions.
This research focuses on addressing these limitations by proposing the design and implementation of a native PREDICT() DAX function. This function would enable Power BI users to load and score ONNX (Open Neural Network Exchange) formatted machine learning models directly within DAX measures and calculated columns. The PREDICT() function is envisioned as a lightweight, efficient, and secure alternative to external ML services, supporting real-time scoring and seamless interaction with report filters and slicers. By embedding ML logic directly into the semantic model, this approach allows business users to derive predictive insights interactively, without external dependencies or complex refresh cycles.
The paper presents a detailed architecture for the proposed function, including its integration within the VertiPaq engine, data flow between the semantic model and ONNX runtime, and sandboxing for security compliance. It also outlines use cases such as demand forecasting, churn prediction, and anomaly detection, demonstrating how native ML scoring in DAX can transform business decision-making. By reducing technical barriers and centralizing analytics within the BI tool, the proposed approach aims to democratize access to predictive modeling and elevate the strategic impact of Power BI across organizations.

Vulnerability Management at Scale: Automated Frameworks for 100K+ Asset Environments

2025-09-09T14:55:03+00:00

Enterprises operating at hyperscale spanning over 100,000 endpoints, servers, and cloud assets face unique challenges in managing vulnerabilities effectively. Traditional vulnerability management (VM) tools and workflows struggle to cope with the volume, velocity, and contextual complexity of findings in such environments. Manual triage, CVSS-only scoring, and siloed remediation processes result in delayed response times, audit failures, and heightened risk exposure. As the attack surface grows dynamically, organizations require scalable, automated solutions that provide continuous visibility, contextual risk prioritization, and orchestrated remediation.
This paper proposes a modular, automation-driven framework for vulnerability management at scale. The architecture integrates continuous asset discovery, threat enrichment, risk-based prioritization, and response automation using tools such as Tenable, ServiceNow, and Cortex XSOAR. It shifts prioritization from static scoring models to contextual models incorporating exploitability, asset criticality, and threat intelligence sources like EPSS and CISA KEV. Evaluations conducted across large enterprises demonstrate a 55% reduction in mean time to remediation (MTTR), a 2.3x improvement in SLA adherence, and a 75% reduction in manual remediation effort. Case studies validate the framework's effectiveness in complex, compliance-driven industries such as healthcare and financial services. The paper concludes with strategic recommendations and future directions involving AI-based risk modeling, SBOM integration, and Zero Trust enforcement. This research offers a repeatable blueprint for security leaders seeking to operationalize vulnerability management in high-scale environments through automation, intelligence, identity and access governance and cross-platform integration.

Multi-Cloud Transaction Orchestration for Hybrid TIBCO/java Micro-services

2025-09-09T14:58:50+00:00

A major cause is likely the rapid dynamism of development of enterprise applications which bring with them the compelling urge for enterprises to design and implement multi-cloud strategies that exploit the different clouds to have the advantages of resilience and scalability as well as cost control. In a well-structured multi-cloud services spectrum, one can gamble that plans to enable transactions ’ distribution,
especially about minimal-functionality and failover resilience, will be insufficient. The project addresses the design and implementation of a Transaction Orchestration Layer for Multi-Clouds that leverages full TIBCO and Java-based microservices, with the following goals:
Supporting the event-driven orchestration, smart routing, and workload sharing of a hybrid environment, alongside supporting multiple cloud vendors under the same ACID and BASE transactional model. The technique draws on the ability of the TIBCO Engine to mix colors using Java based micro-system and such a combination facilitates transaction coordination services together with the discovery and monitoring of services in hybrid cloud environments. The testing has demonstrated that the approach decreases the queue of transactions, reduces latency and improves resistance to faults. Consequently, it has the potential of use in multi-cloud enterprise systems.

Choosing the Right Data Deployment Architecture in Industry 4.0: From Sensors to Decisions

2025-09-09T15:03:22+00:00

The future of manufacturing is not about producing more; it is about producing smarter and more efficiently. Industry 4.0 combines advanced technologies such as the Internet of Things (IoT), artificial intelligence (AI), and big data into manufacturing and supply chain practices. Modern machines come with smart sensors and connected systems that enable the acquisition of data that can be used to inform a wide variety of data-driven decisions, such as predictive maintenance, quality assurance, optimizing energy use, and production governing. However, the value of this data is not apparent until it has been properly processed and analyzed. Analytics, machine learning, and real-time processing frameworks transform the data into information and provide actionable decision-making ability so that firms can be forewarned of future failure, optimize resource consumption, and respond proactively to dynamic challenges of an industrial environment. This paper presents a comparative study of four primary approaches to data deployment architecture in Industry 4.0: on-prem, near-prem (edge), cloud, and hybrid, highlighting their strengths and weaknesses. Latency requirements, security, cost, scalability, and regulatory compliance are often determining constraints to help the industry choose one of these approaches. The study contributes by offering a structured evaluation of data deployment architectures and providing practical guidelines to align architectural choices with industrial needs, thereby supporting more effective adoption of Industry 4.0 strategies.

Quantitative Analysis of Cost-Benefit Models for PCI DSS Control Implementation in Financial Organizations for Risk-Optimized Investment Decisions

2025-09-09T15:05:20+00:00

The introduction of Payment Card Industry Data Security Standard (PCI DSS) version 4.0 has established 64 new requirements, thereby complicating organizations' strategies regarding cybersecurity investment decisions. Data breach costs average $4.88 million globally, leading organizations to seek strategic optimization of compliance investments instead of perceiving them solely as obligatory expenses. This study identifies a significant gap by creating a detailed quantitative framework to optimize cost-benefit decisions in implementing PCI DSS controls across various organizational settings. This study employs quantitative modeling informed by the Gordon-Loeb Model, integrating contemporary threat intelligence from the Verizon Data Breach Investigations Report 2025, breach cost data from the IBM Security Cost of a Data Breach Report 2025, and implementation benchmarks to develop systematic decision support tools. The framework underwent validation via case studies encompassing small, medium, and large enterprises that exemplify standard PCI DSS compliance scenarios. The analysis indicates that investments in PCI DSS compliance yield significant positive returns, with ROI ranging from 21% to 1,107%, and most investments experiencing payback periods between 0.2 and 1.5 years. The development of an information security policy represents the most significant initial investment. Small businesses exhibit remarkable ROI profiles ranging from 504% to 1,107%, medium organizations attain steady positive returns between 21% and 278%, and large enterprises experience significant absolute risk reduction, albeit with more extended payback periods. The study primarily integrates theoretical cybersecurity investment models with practical compliance optimization. It offers organizations empirically-based frameworks for prioritizing security investments while meeting compliance goals and enhancing business value.

Machine Learning-Based Intrusion Detection System

2025-09-09T15:07:36+00:00

An explanation of IDSs is given in this work. Recent developments in technology have sparked worries about privacy and security. Network security is becoming more and more crucial as cyber networks and their uses grow. Intrusion detection systems (IDSs) that use machine learning are successful; in particular, the Supervised Model raises detection rates. People may find it challenging to comprehend their choices when faced with complex models. The majority of current model interpretation research is concentrated in domains including biology, computer vision, and natural language processing. Experts in cybersecurity find it difficult to maximize choices based on model evaluations in real life. A framework is proposed to handle these issues.

Deep Learning-Enhanced Hybrid AI and Zero Trust Framework for Secure Cloud-Centric Software Development

2025-09-10T05:46:35+00:00

As organizations embrace cloud technologies and software-driven operations, they gradually become targets for adversaries who take advantage of intricate supply chains, miss configurations, and the proliferation of identities. Traditional, reactive defenses that focus on perimeter security struggle to counter advanced persistent threats, ransom ware attacks, and breaches in software supply chains. This paper introduces a hybrid cyber security framework that integrates Artificial Intelligence (AI), deep learning, Zero Trust Architecture (ZTA), and cloud-native security practices to safeguard the software development lifecycle (SDLC) comprehensively. The framework offers several key benefits: (i) AI-enhanced threat intelligence and risk assessment at every stage from code creation to deployment and runtime; (ii) deep learning-supported integrity checks for CI/CD provenance and verifiable change management; (iii) continuous authentication and least-privilege authorization based on Zero Trust principles; and (iv) cloud security posture management that aligns with shared responsibility models. We outline the threat model, architecture, data flows, and a detailed implementation plan. Our evaluation employs a mixed-methods approach, incorporating a systematic literature review, industry surveys, and a case study in the healthcare sector. We also correlate our controls with top industry standards, including NIST SP 800-207/53/218, CSA CCM, and ISO/IEC 27001. The results show significant improvements in detection accuracy, a reduction in mean time to respond (MTTR), robust tamper-evident pipelines, and lower risks of lateral movement. Our contributions include a reference architecture, a maturity model, and compliance mapping designed for secure cloud-centric development.

Automated Detection and Characterization of Kidney Cysts with Deep Learning from Ultrasonographic Images

2025-09-10T05:58:57+00:00

Ultrasonography is frequently used to identify disorders affecting internal organs since it is non-invasive, non-radioactive, real-time, and reasonably priced. To measure organs and tumours in ultrasonography, a set of measurement markers is positioned at two different places. The size and location of the target finding are then determined using this information. Renal cysts, which affect 20–50% of patients regardless of age, are one of the measurement objectives of abdominal ultrasonography. Because kidney cysts are frequently measured in ultrasound imaging, automating the measurement would also have a significant influence. Creating a deep learning algorithm that can quickly detect kidney cysts in ultrasound pictures and forecast the positions of two crucial anatomical markers to gauge the cysts' sizes was the aim of this study. The deep learning method used refined UNet++ to forecast saliency maps, which show the positions of important landmarks, and refined YOLOv5 to detect kidney cysts. After identifying the ultrasound images, YOLOv5 clipped them into the bounding box and sent them to UNet++. Three sonographers visually positioned important landmarks on 100 test data items that were not visible in order to compare the outcomes with human performance. A board-certified radiologist said that these well-known, famous locations provided the ground truth. Next, we assessed and contrasted the deep learning model's and the sonographers' accuracy. Measurement error and precision-recall metrics were used to assess their performances. The analysis's findings demonstrate that our deep learning model's accuracy and recall for identifying renal cysts are on par with those of conventional radiologists; the radiologists' accuracy and speed in predicting the crucial landmark placements were on par with ours.

AN INTELLIGENT HYBRID SYSTEM FRAMEWORK FOR CARDIOVASCULAR DISEASE PREDICTION EMPLOYING MACHINE LEARNING TECHNIQUES

2025-09-10T06:02:17+00:00

One of the most deadly diseases in the world, heart disease significantly affects people's life. It makes reference to the heart's inability to sufficiently pump blood to various body parts. Both preventing and treating heart failure depend on an accurate and timely identification of cardiac illness. It has long been believed that medical histories are unreliable for diagnosing heart issues. However, noninvasive techniques such as machine learning have proven to be reliable and effective in distinguishing between those with heart disease and those in good health. Using a dataset on heart illness, we created a machine learning-based diagnostic system for heart disease prediction in this proposed research. Seven well-known machine learning algorithms, three feature selection methods, seven classifier performance evaluation criteria, and a cross-validation approach were all employed. Among these requirements are execution speed, specificity, sensitivity, Matthews' correlation coefficient, and classification accuracy. Our proposed approach successfully distinguishes and categorises individuals with heart disease from healthy individuals. We also calculated the receiver operating characteristic curves and the area under the curves for each classifier. This paper covers in detail all classifiers, feature selection algorithms, preprocessing techniques, validation techniques, and metrics for evaluating the performance of the classifiers. Three feature selection methods, seven well-known machine learning algorithms, and a cross-validation approach were employed. In addition, we evaluated the classifiers' performance using seven criteria: classification accuracy, specificity, sensitivity, execution time, and Matthews' correlation coefficient. The proposed method effectively identifies and categorises people with heart disease when compared to healthy individuals. We also calculated the receiver operating characteristic curves and the area under the curves for each classifier. This paper covers all classifiers, feature selection algorithms, preprocessing techniques, validation processes, and metrics for evaluating classifier performance in detail.

AUM block geometric mean labeling for tadpole, kayak paddle and bull graphs

2025-09-11T05:55:29+00:00

Graph labeling is an interesting area of mathematics with many real-world uses, such as in medical coding, networking, transportation, disease diagnosis, and cryptography. The AUM block geometric mean labeling is a recently introduced labeling. In this paper we obtain AUM block geometric mean labeling for tadpole, kayak paddle, and bull graphs. Suitable examples are given.

Emotion and Drowsiness Detection System using Multimodal Fusion and Explainable AI

2025-09-15T06:07:35+00:00

This paper illustrates the creation of a holistic Real- Time Driver Emotion and Drowsiness Detection System that is supposed to enhance road safety through detection of a driver emotional state and the level of alertness. The application integrates the process of analysis of facial expressions, speech emotion recognition, and drowsiness detection to deliver real-time interventions and make drivers safe and aware of driving in the streets. Based on multimodal AI methods, the system both reads visual and audio information, as well as reads facial expression and speech to identify emotional states and monitors faces to recognize the signs of drowsiness. The system puts more focus on the detection of drowsiness as opposed to emotion recognition to facilitate the provision of safety alerts in time. Also, the Explainable AI (XAI) integration brings transparency in the sense that the users can see how the system makes decisions. This project emphasizes the significance of multimodal fusion, the fact is that both results of facial and speech analysis are used simultaneously to provide better and more objective results, therefore, enhancing the overall safety of drivers. The proposed system can be deployed to be in use in the real-world context with little disruption to the user experience giving the feedback in real-time and ensures safety is monitored throughout driving.

Proactive Cost Governance through Automated FinOps Platforms: A Platform Engineering Perspective for Financial Services

2025-09-16T06:13:35+00:00

Financial services organizations face unprecedented challenges in managing cloud infrastructure

costs while maintaining regulatory compliance. Traditional reactive cost monitoring fails to prevent

budget overruns and compliance violations before they occur. This article presents a

comprehensive framework for engineering automated FinOps platforms that embed financial

intelligence directly into infrastructure lifecycles. The proposed architecture transforms cost

management from post-deployment reporting to pre-deployment prevention through sophisticated

CI/CD pipeline integration, where cost functions as a quality gate alongside security and

performance metrics.

b – Coloring of Extended Duplication Graph of Dragon graph(Tn,1) and Sunlet graph(Sn) Networks

2025-09-16T06:24:08+00:00

The b – chromatic number ????(????), is the highest k for which G confess a b-coloring by k colors. A b-coloring of G by k colors is proper vertex k-coloring such that in each color class i there is a vertex ????_???? having neighbors in every other k − 1 color classes. In this article we obtain b – chromatic number of (M[EDG(Tn,1)]) middle graph of extended duplication graph of Dragon graph, middle graph of extended duplicate graph of Sunlet graph (M[EDG(Sn)]), and total graph of (T[EDG(Tn,1)]) extended duplication graph of Dragon graph and total graph of extended duplicate graph of Sunlet graph T[EDG(Sn)].

LAPLACIAN SPECTRUM IN STEREOPHONIC SYSTEMS

2025-09-16T06:29:04+00:00

Contemporary innovations are deeply intertwined with mathematics, which provides the foundational tools for modeling and problem-solving. Real-world challenges can be effectively represented through graph-theoretic frameworks, enabling structured analysis and solution design. In the context of emerging electric vehicle technologies, the customization of stereophonic systems for enhancing travel experiences can be formulated using fuzzy graphs. Furthermore, this paper emphasizes on the application of Laplacian spectrum of complete fuzzy graphs for optimizing system performance.

THE PROSPECT OF ARTIFICIAL INTELLIGENCE- BASED WOOD SURFACE INSPECTION

2025-09-17T10:59:48+00:00

Identifying wood errors is an essential part of the production process because the woodworking business is very dependent on the final wood products that meet quality standards. This has always been a labor -intensive process that requires experienced individuals physically verify each element. But new opportunities for automatic and improvement to detect wood errors have emerged as intensive learning algorithms have improved.

Spectral Properties of Operators Satisfying the Generalized Weyl– Drazin Property with Applications to Machine Learning

2025-09-18T04:48:57+00:00

In this paper, we investigate the spectral properties of bounded linear operators that

satisfy the generalized Weyl–Drazin (ℊ????????) property. We establish new inclusions and

relationships between the Drazin spectrum, poles of the resolvent, and the Weyl

spectrum. Several supporting lemmas, examples, and corollaries are presented to

illustrate the theoretical framework.

Minimum Tree t-Spanners on Fuzzy Ratio Labelled General Graphs - NP Complete

2025-09-18T05:57:23+00:00

Tree t-spanners are trees that span a graph that involve distances and are useful in

designing telecommunication, electrical networks, civil network planning, and more. Extending the

idea of tree t–spanner in fuzzy graphs is the concept of this paper. The study focuses on tree t

spanners of ratio-labelled fuzzy graphs using the Breadth-First Search (BFS) algorithm, which is

crucial for optimizing communication and routing efficiency in networks

AI-Powered Smart Water Distribution System: An Intelligent Approach for Resource Optimization

2025-09-18T08:56:50+00:00

Water distribution systems face critical challenges

including leakage detection, demand prediction, and

optimization of resource allocation. This research

presents a novel AI-powered smart water distribution

system that integrates machine learning algorithms,

IoT sensors, and cloud computing to revolutionize

water management. The proposed system employs a

hybrid approach combining deep learning for demand

forecasting, reinforcement learning for valve control

optimization, and anomaly detection algorithms for

leak identification. Experimental results demonstrate

significant improvements in water conservation

(28%), operational cost reduction (32%), and leak

detection accuracy (94.7%) compared to conventional

systems. The framework's scalable architecture allows

for seamless implementation across various urban

water infrastructures, offering a sustainable solution

to global water management challenges.

Classification of Diabetic Retinopathy Disease Levels by Extracting Topological Features Using Graph Neural Networks

2025-09-19T04:53:12+00:00

Diabetic retinopathy (DR) is one of

the most common causes of blindness in the

world”. It has to be diagnosed quickly and

accurately so that treatment may begin as soon

as possible. Manual examination of fundus

photographs by doctors is prone to mistakes

and is labor-intensive. “Using computer-assisted

methods, especially Convolutional Neural

Networks (CNNs), to automate DR diagnosis

seems promising”.

An Approach to Solution of Fractional Kinetic Equations Associated with Hyper-Bessel Function and k-Mittag-Leffeler Function

2025-09-20T06:38:54+00:00

In this present work, expressed the generalized fractional kinetic equations associated with Hyper

Bessel function and their fractional derivatives. Additionally, solutions are derived in terms of the

Mittag-Leffler function by employing the Laplace transform approach. Furthermore, a graphical

depiction of the solutions is offered to discuss particular cases of fractional kinetic equations and

show how they behave.

A NOVEL APPROACH TO FRACTIONAL KINETIC EQUATIONS INVOLVING LAGUERRE POLYNOMIAL FUNCTION AND S-FUNCTION

2025-09-20T07:00:33+00:00

Fractional kinetic equations incorporating special functions have proven useful in explaining and

solving many significant mathematical and mathematical physics problems. Given the significant

role of arbitrary-order kinetic equations, this study focuses on solving a newly formulated equation

of this type by utilizing the Sumudu transform. The equation incorporates fractional derivatives and

involves a composition of Laguerre polynomials and the S-Function. Our investigation included

MATLAB-generated graphical representations to show how these solutions behave under different

parametric conditions. It is important to highlight that the study's results are incredibly flexible and

could lead to confirmed and perhaps undiscovered research findings in this area.

DETECTIONOF EMOTIONS INCATEGORICAL TWEETS USING NAIVE BAYES AND SVM ALGORITHM

2025-09-20T07:04:08+00:00

In this project, we developed an emotion detection

framework designed to identify emotional cues in tweets. Emotions

play a central role in our daily lives, and with the widespread use of

social media, platforms like Twitter have become a valuable source

for understanding people’s feelings and opinions. Users express

themselves in many ways—some share positive thoughts, while

others may post harmful or bullying content. By analyzing these

tweets, we can gain insights into public opinion on news, events, and

social issues.

Air Quality Index Forecasting via Genetic AlgorithmBased Improved Extreme Learning Machine

2025-09-20T07:08:12+00:00

Air quality prediction plays a

vital role in safeguarding public health

and guiding environmental policy.

Traditional single-model approaches

often struggle to accurately forecast air

quality fluctuations. In response, this

study introduces a robust prediction

system leveraging advanced machine

learning techniques. We present a

comparative analysis of several models

including Support Vector Regression

(SVR), Genetic Algorithm-Enhanced

Extreme Learning Machine (GA-KELM),

and Deep Belief Network with Back

Propagation (DBN-BP). Additionally, we

propose the integration of Bidirectional

Long Short-Term Memory (BiLSTM), a

deep learning architecture, to further

enhance prediction accuracy.

A Generative AI-Driven Information System for Behavioral Detection of Zero- Day Cyber Attacks

2025-09-20T07:10:49+00:00

Zero-day attacks are among the most serious cybersecurity threats because they

occur before recognizable signatures are available, rendering traditional detection

methods ineffective. This research aims to develop an intelligent, generative AI

based detection framework capable of simulating and identifying unknown cyber

threats in real time. In this research paper, we present a model that uses generative

transformers to simulate sophisticated attack behaviors based on historical

sequence data and attacker profiles.

Enhanced Quantum Key Distribution Protocol to Improve Robustness and Error Minimization

2025-09-20T07:15:28+00:00

Quantum cryptography represents a swiftly advancing discipline with the capacity to
fundamentally transform the landscape of data security. A prime application in this
domain is quantum key distribution (QKD), which facilitates secure communication by
leveraging the principles of quantum mechanics to generate and disseminate
cryptographic keys. This study is concentrated on the methodologies for safeguarding
communication and data transfer against cyber-attacks through the utilization of
quantum cryptography, particularly focusing on the quantum key distribution protocol
known as BB84. BB84 is the first protocol for QKD in the year 1984. In this paper we
simulated a new way to enhance quantum key distribution protocol BB84 by introducing
multi-basis encoding, enhanced error correction, and built-in authentication. The
additional basis Y provides increased robustness against intercept resend attacks, and the
authentication mechanism ensures secure classical communication and hence there will
be no loss of information.

TEACHING THE TOPIC “THE EFFECT OF RADIATION ON THE BODY” BY USING THE “DISCUSSION” METHOD

2025-09-23T16:12:46+00:00

This article presents the results of a scientific study on "The Effect of Radiation on the Body," using the "Discussion" method. Data, achievements, and shortcomings that help comprehensively illuminate the issue using modern methods were studied. Pedagogical analysis was conducted along with correct solutions to posed tasks and tests, and innovative ways to facilitate better comprehension by students were discussed. The "Discussion" method demonstrated an efficiency of 12% in the teaching process.

constraints sorting method’ new approach for solving large-scale problems

2025-09-25T06:37:14+00:00

In this paper, we propose an iterative method for solving very large-scale linear problems called the constraints sorting method(CSM), it consists in sorting the constraints of the initial problem, and iteratively solving a serie of sub-problems of in- creasing size which will converge to the solution sought, the efficiency of this method depends on the choice of constraints to be introduced, we have chosen to add, at each it- eration, a set of constraints most orthogonal to the criterion of the problem to be solved, which gave us very good results. In order to compare (CSM) with the interior point method, we have realized a numerical implementation of our (CSM) approach using the Matlab programming language, and numerical results on execution time showing that our approach is competitive are presented.

Implementation and Validation of a Framework for an Ethical and Cyber-secure Smart City

2025-09-25T06:39:29+00:00

In recent years, smart cities have experienced a transformative shift aimed at improving the quality of life for both residents and the broader community. Our research introduces a novel framework that enhances data secrecy and authentication to bolster security in smart city. This paper synthesizes recent advancements in Stochastic Multicriteria Decision Making (SMCDM) and Decision Support Systems (DSS) to address these critical issues. It explores frameworks that model evaluations as random variables, providing a more robust analysis than traditional deterministic approaches. A central focus is the Stochastic Multicriteria Acceptability Analysis (SMAA) family of methods, which operates by exploring the weight space to identify preferences supporting each alternative, thereby circumventing the need for explicit preference elicitation. Key findings highlight SMAA's ability to provide descriptive insights through acceptability indices, central weight vectors, and confidence factors. The synthesis also emphasizes the crucial impact of accounting for dependent uncertainties, demonstrating how their neglect significantly weakens decision support. These methods collectively offer a powerful paradigm for enhancing decision quality in uncertain and group-oriented contexts. The rapid development of smart cities introduces significant ethical and cybersecurity challenges, including data privacy violations, algorithmic bias, and vulnerability to cyberattacks. This paper proposes a novel fuzzy-based multi-criteria decision-making (MCDM) framework for designing, implementing, and validating an ethical and cyber-secure smart city model. The framework integrates fuzzy logic to handle uncertainty in expert evaluations and optimizes the selection of smart city architectures based on ethical compliance and cybersecurity robustness. We evaluate five alternative smart city frameworks using Fuzzy TOPSIS and validate the results through sensitivity analysis, comparative AHP assessment, and real-world case studies.

Enhancing Security and Real-Time Resilience in Microservices: Automated Self-Healing Controls for Secure gRPC over HTTP/3 with AES-256-GCM

2025-09-25T06:41:39+00:00

Microservices are essential to modern distributed systems because they allow for scalability and agility, but they also present significant security and performance issues that call for rigorous mathematical answers. In this study, a performance–security model is used to describe an integrated framework for secure gRPC communication over HTTP/3 with AES-256-GCM encryption. We derive end-to-end latency analytical expressions Ltotal = Lnet + Lenc + Lproc and demonstrate how QUIC's multiplexing and 0-RTT handshake reduce Lnet whereas hardware-accelerated AES lowers Lenc We provide an algorithmic specification for runtime anomaly detection and automatic remediation, formally characterize the threat model of the system, demonstrate its termination, and bound the remediation time complexity to O(n) with regard to the number of microservices. The model's experimental deployment on a 5-node Kubernetes cluster verifies it: with only 5% more CPU overhead, mean latency dropped by 23% and throughput increased by 20% when compared to baseline HTTP/2 + TLS 1.2. The suggested approach shows how provably safe, robust, and high-performance microservice architectures can be achieved by mathematically based design of anomaly detection, cryptographic controls, and self-healing mechanisms.

Leveraging Speaker Embeddings from Speaker Verification for Controllable Multispeaker Text-to-Speech

2025-09-26T05:28:11+00:00

Our hybrid system is compatible with Librosa library, Gaussian Mixture Model, and text-to-speech (TTS) technology. Neural networks are then used by TTS to produce audio speech that mimics the sounds of several speakers, including those that are excluded from the training set. The system incorporates three separately trained components: (1) With a small clip of the target speaker audio, the pre-trained encoder can validate it after comparing with a stand alone separately stored dataset of thousands of speakers' high pitched vocal notes without transcripts can produce fixed-length embedding vectors; (2) TacotronII endorsed sequential model that, relies on the primary level speaker embedding, transforms text into mel-spectrograms; (3) An auto regressive ‘WaveNet vocoder’ converts Mel spectrograms to waveforms with are functions of time.We demonstrate how the discriminative pre-training of the speech encoder on large-scale speaker diversity conveys important knowledge about speaker variability to the multi-speaker TTS challenge, allowing high-quality synthesis even for unknown speakers. We measure the advantages of rich and heterogeneous speaker datasets for enhanced generalization. Progressively, the new sounds generated with the aid of embedding of a random/ variable speaker can effectively generate new sounds that are different from the training set, suggesting that the model has picked up strong speaker representations.To prevent undue similarities, alternate wording and structure are used while preserving the essential factual data. Our recently added custom Librosa layers extract necessary features, which is helpful to improve the efficiency of the particular feature, and our freshly added custom GMM layers eliminate noise from the raw audios by removing noisy features.

E-Learning Effectiveness and Efficiency in Kassala and Gedaref Universities: An IS-Impact Evaluation

2025-09-26T10:12:53+00:00

In low-resource settings with intermittent power and connectivity, e-learning must translate technical and informational qualities into institutional value. This cross-sectional study evaluates how e-learning creates organizational impact in two Sudanese public universities (Kassala, Gedaref) using the IS-Impact model. Survey data from N = 412 students and faculty were analyzed with covariance-based SEM (WLSMV). Model fit was good. System Quality (SQ) strongly predicted Information Quality (IQ), and both IQ and SQ improved Individual Impact (II); in turn, II strongly drove Organizational Impact (OI). Direct paths from SQ/IQ to OI were non-significant, while indirect effects via II were substantial (e.g., IQ→II→OI = 0.41; SQ→II→OI = 0.16), indicating a full/semi-full mediation pattern. Connectivity and electricity reliability amplified quality→II links, underscoring the importance of engineering for uptime and low-bandwidth delivery. Results were invariant across roles and universities. Implication: institutions should prioritize editorial pipelines that raise IQ, harden SQ for mobile/low bandwidth, and invest in reliability measures (offline caching, micro-servers, basic power backup), monitored through actionable KPIs. Institutional approvals were obtained. Index Terms— IS-Impact; E-learning; System Quality; Information Quality; Low bandwidth; Sudan.

MORPHOLOGICAL, OPTICAL AND CYCLIC VOLTAMMETRY STUDIES ON HAFNIUM OXIDE NANOPARTICLES

2025-09-26T10:31:54+00:00

Hafnium oxide nanoparticles were synthesized through the microwave-assisted chemical method. The as-prepared samples were calcined at 700 °C. The synthesized hafnium oxide nanoparticles were characterized by Scanning Electron Microscope with energy dispersive spectroscopy and UV-visible spectroscopy. The scanning electron micrographs of the materials showed uniform distribution of particles. Energy dispersive spectroscopy confirms the purity of the prepared sample. UV-Visible spectrum hafnium oxide nanoparticles showed the minimum absorption in the entire visible region. The energy bandgap of the synthesized hafnium oxide nanoparticles was determined through Tauc plot and found to be 5.65 eV. The cyclic voltagram of the hafnium oxide nanoparticles were studied through electrochemical impedance analyzer. The specific capacitance and energy density of hafnium oxide nanoparticles were determined.

Efficacy Analysis of Fractured Bone on Hyper Spectral Imagery

2025-09-26T15:00:10+00:00

Objectives: Detection of Fractures: Assess the ability of hyper spectral imagery to accurately detect fractures in bone structures. This involves identifying spectral signatures associated with fractured bone and developing algorithms for automated detection. Classification of Fracture Types: Different types of fractures (e.g., hairline fractures, compound fractures) may exhibit distinct spectral characteristics. Objectives should include classifying fracture types based on hyper spectral data to aid in diagnosis and treatment planning. Method: Data Acquisition: Acquire hyper spectral imagery data containing spectral information across a wide range of wavelengths. This data can be obtained from specialized hyper spectral imaging systems capable of capturing spectral signatures of the target area. Preprocessing: Correct for atmospheric effects and sensor artifacts to enhance the quality of the hyper spectral data. Perform radiometric calibration to convert raw spectral data into quantitative reflectance or absorbance values. Remove noise and irrelevant spectral bands to focus on informative spectral features relevant to fractured bone. Region of Interest (ROI) Selection: Identify the region of interest containing the fractured bone within the hyper spectral image. Define the boundaries of the fractured area and ensure that the ROI encompasses relevant spectral information. Findings: Detection Accuracy: The analysis may reveal the ability of hyper spectral imagery to accurately detect fractures in bone structures. This includes the sensitivity and specificity of fracture detection compared to ground truth data or conventional imaging methods. Classification of Fracture Types: Hyper spectral imaging may enable the differentiation of different types of fractures based on their spectral signatures. Findings may include the ability to distinguish between hairline fractures, compound fractures, and other fracture types. Quantification of Fracture Severity: The study may demonstrate the efficacy of hyper spectral imaging in quantifying fracture severity parameters such as size, displacement, and fragmentation. This could provide valuable information for clinical decision-making and treatment planning. Novelty: Non-Invasive Imaging: Hyper spectral imaging offers a non-invasive means of assessing fractures, unlike traditional methods such as X-rays or CT scans, which involve ionizing radiation. This reduces patient exposure to harmful radiation and provides a safer alternative for imaging fractures, especially in pediatric or pregnant populations. Comprehensive Spectral Information: Hyper spectral imagery captures spectral information across a wide range of wavelengths, providing a rich dataset for analysis. Unlike conventional imaging modalities that rely on anatomical structures, hyper spectral imaging exploits biochemical and molecular signatures, offering insights into tissue composition and pathology.

LAGRANGE CONDITIONAL GENERATIVE ADVERSARIAL NETWORK (LCGAN) BASED IMAGE DENOISING AND MULTI-SCALE DUAL ATTENTION INCEPTION V3 (MSDAIV3) FOR TOMATO LEAF DISEASES DETECTION

2025-09-26T15:02:43+00:00

Agriculture has been the primary source of income for the majority of people in India. The impact of plant illness ranges from mild manifestation to the destruction of complete plantations that severely affect the agricultural economy. Therefore, early detection and diagnosis of these diseases are essential. During diagnosis, noise can be easily generated during image acquisition, transmission, and processing, which makes it challenging to extract disease features and reduces diagnosis accuracy. In this paper, new identification model is introduced for tomato leaf disease. Firstly, a Lagrange Conditional Generative Adversarial Network (LCGAN) is introduced to reduce the image noise interference. LCGAN, noise removed image is estimated by an end-to-end trainable neural network. LCGAN includes of encoder and decoder architecture so that it can generate better detection results. LCGAN is used to decrease the difficulty of extracting tomato leaf disease features in the identification network. Secondly, three-segment linear conversion is utilized to increase image contrast. Thirdly, Inception-V3 is proposed to extract abundant disease features. In Inception-v3, a batch normalization (BN) layer is inserted as a regularizer between the auxiliary classifier and the fully connected layer. Finally, Multi-Scale Dual Attention Inception V3 (MSDAIV3) model is introduced to measure the inter-class similarity and intra-class variability identification of tomato leaf diseases. The proposed model has been trained and tested extensively on real time dataset with six classes (Bacterial spot, early blight, Leaf Mold, Septoria leaf spot, Leaf Curl Virus, and Healthy). Denoising results of proposed model is compared to existing methods using metrics such as Peak Signal-to-Noise Ratio (PSNR), Structural Similarity Index (SSIM), and Feature SIMilarity Index for Color images (FSIMc). The proposed approach shows its superiority over the existing methods using metrics like precision, recall, f-measure, and accuracy.

A New form of N- Hexa Topological Approach to Reducing Maternal Anxiety After Delivery

2025-09-27T06:37:01+00:00

Anxiety during pregnancy is associated with adverse outcomes in mothers and infants. Unfortunately, as anxiety is often synonymously mentioned with depression, the studies focusing solely on anxiety during pregnancy are not as robust as those in the field of depression are. This study employs the N-Hexa Topology framework to identify critical factors influencing maternal and newborn care outcomes. Six key domains were analyzed, each encompassing rel- evant maternal health variables, Early Childhood Health Service, Breastfeeding, Whole Body Massage (F1), Dietary Advice and Routine Antibiotic Prophylaxis (F2), Management of Post- partum Depression and Gestational Weight Gain (F3), Nutritional Interventions and Physical Activity (F4), Newborn Assessment (F5), and Maternal Assessment (F6). Using Nano topology Analysis, core attributes within these factors were isolated based on their impact on maternal health outcomes.

Mathematical Optimization of AI-Based Document Processing Workflows Using Markov Decision Processes

2025-09-27T06:39:40+00:00

Conventional models tend to malfunction with stochastic task arrivals, indefinite processing times, and priority conflicts. The study presents a scalable and wise optimization model based on Markov Decision Process (MDP) along with Deep Reinforcement Learning (DRL) as Proximal Policy Optimization (PPO) to acquire optimal policies for document routing. The OCR Receipts Dataset from Kaggle (2023) is used that contains annotated receipt images. Major techniques involve feature vector encoding, filtering out noise, and balance in reward of latency, accuracy, and cost. Python implementation involves state-action modeling and multi-objective optimization to enhance task scheduling, usage of resources, and decision-making. There is improved accuracy, less processing latency, and adaptive performance compared to baseline models. The framework supports scalable automation of finance, law, and the public sector. On top of that, the PPO agent has strong learning abilities across varying workflow circumstances. The suggested system is up-scalable to other areas which need an intelligent, real-time processing of documents and routing tasks.

Dynamic Pricing in the Cloud Era: How Agentic AI Can Reinvigorate Private Cloud Providers

2025-09-27T06:41:09+00:00

This paper introduces a novel application of agentic AI to address competitive challenges faced by private cloud providers. By integrating real-time competitive intelligence, dynamic pricing adjustments, and automated proposal generation, the proposed AI solution overcomes limitations in current sales methodologies, which struggle to match the agility of hyperscale providers’ pricing models. The system acts as a sales force multiplier, enabling private cloud providers to target cost-conscious mid-market buyers effectively. We detail the AI’s core functionalities, its integration into sales workflows, and its economic benefits, including enhanced sales volume, optimized infrastructure utilization including the Datacenters, and improved profitability. This approach strengthens private cloud providers’ competitive positioning and captures a larger share of the cost-conscious mid-market segment.

DEEP LEARNING-DRIVEN MRI-BASED BRAIN TUMOR DETECTION AND GRADING: A CRITICAL ASSESSMENT AND PERFORMANCE ANALYSIS

2025-09-30T05:10:42+00:00

Brain tumors, both benign and malignant, pose significant health challenges, with early and accurate diagnosis being critical for effective treatment planning. Magnetic Resonance Imaging (MRI) remains the gold standard for non-invasive tumor assessment due to its high soft-tissue contrast and multiparametric capabilities. In recent years, deep learning (DL) techniques have revolutionized brain tumor detection, segmentation, and grading by surpassing traditional methods in accuracy and robustness. This survey provides a detailed review and comparative performance analysis of state-of-the-art DL and machine learning (ML) methods developed between 2022 and 2025. The study examines benchmark MRI datasets such as BraTS, TCGA, REMBRANDT, and Figshare, highlighting their relevance to clinical research. Comparative evaluation reveals the evolution from CNN-based architectures to hybrid CNN–Transformer frameworks and ensemble strategies, with recent models achieving near-perfect classification accuracy. Despite these advancements, challenges remain regarding model interpretability, computational efficiency, and generalization to heterogeneous clinical data. The assessment concludes by identifying research gaps and outlining future directions for lightweight, explainable, and clinically deployable AI solutions in neuro-oncology.

Implementing Incremental Data Loading Strategies for Efficient Cloud Migrations

2025-09-30T05:12:38+00:00

This comprehensive article explores incremental data loading strategies as an essential approach for efficient cloud migrations in enterprise environments. By dividing massive data transfers into manageable segments, organizations can overcome the significant challenges associated with traditional "big bang" migration methods. The article examines how incremental loading minimizes operational disruptions while optimizing resource utilization across diverse cloud platforms. It analyzes key implementation techniques, including Change Data Capture (CDC), timestamp-based filtering, and partition-based migration, each offering distinct advantages depending on specific organizational requirements. The article further explores robust error handling mechanisms that dramatically reduce recovery times and enhance data integrity throughout the migration process. Through a detailed case study of BigQuery to Redshift migrations, the article demonstrates how incremental approaches enable organizations to maintain business continuity while methodically transitioning between cloud environments. This sophisticated methodology aligns perfectly with modern enterprise requirements for high availability and operational resilience, transforming potentially disruptive migration projects into predictable, low-risk transitions that maximize the benefits of cloud adoption.

Twain Secure Perfect Dominating Sets and Twain Secure Perfect Domination Polynomials of Centipedes

2025-09-30T05:14:26+00:00

Let ???? = (????,????) be a simple graph. A set ???? ⊆ ???? is a dominating set of ????, if every vertex in ???? \???? is adjacent to at least one vertex in ????. A subset ???? of ???? is called a twain secure perfect dominating set of ???? (TSPD-set) if every vertex ???? ∈ ???? \???? is adjacent to exactly one vertex ???? ∈ ???? and (????\{????}) ∪ {????} is a dominating set of ????. The twain secure perfect domination number of ????, represented as ????????????????(????), is the lowest cardinality of a twain secure perfect dominating set of ????. The centipede with 2???? vertices is represented by ????????∗, and the family of all twain secure perfect dominating sets of ????????∗ with cardinality ???? is represented by ????????????????(????????∗ ,????). Let ????????????????(????????∗ ,????)=|????????????????(????????∗ ,????)|. This article builds ????????????????(????????∗ ,????) and derives a recursive formula for ????????????????(????????∗ ,????). With this recursive formula, we examine the polynomial ????????????????(????????∗ ,????)=Σ????????????????(????????∗ ,????)2????????=???????????? which we call the twain secure perfect domination polynomial of centipedes. To create all twain secure perfect dominating sets of centipedes and twain secure perfect domination polynomials of centipedes, we employ a recursive approach in this study.

Challenges faced by the university students and their cognitive skills in learning basic mathematics

2025-10-03T07:11:54+00:00

This study recognized the relationship between the students’cognitive skills and challenges in learning basic mathematics with all first-year students in Cebu Technological University Tuburan -Campus. Employ,using modified survey questionnaires among the 304 first -year students, most are aged 19-20 (73.03%)and have completed 13-14 years of schooling.Financially, 73.68% of their parentsfall into a low- income, with varied educational backgrounds, predominantly at the high school and elementary levels. Math assessments show that 40.13% of students scored above average in “ Number Sense,” 19.14% in “Problem- Solving,”23.36% in “Seriation,” and 52.63% in “Logical Multiplication.” Key challenges include short retention spans (mean =2.47), inattentiveness (mean =2.16), and insufficient lesson planning (mean =2.26). School – related issues, such as a lack of educational materials (mean = 2.30), also impact learning. Significant correlations were found between these challenges and cognitive skills, particularly in problem – solving (r = -.352, p<0.01), seriation (r = -.348, p<0.01), and logical multiplication (r= -.359, p<0.01). Gender and age also play key roles, with notable correlations between family- related challenges and gender (r = -0.220, p<0.01), and between cognitive skills and age (r = -.250 to -.407, p<0.01). The findings revealed that most students found difficult in answering towards problem – solving. This suggests that learning enhancement activity program should be implemented to facilitate the students’ cognitive skills, particularly problem- solving, and to address the challenges affecting their mathematical performance. Such interventions are essential for improving academic outcomes and ensuring that students are better prepared to tackle mathematical concepts.

Enhancing Brain MRI Classification Through Layer-Wise Transfer Learning and Deep Fine-Tuning

2025-10-04T07:44:13+00:00

Brain tumors must be correctly and quickly diagnosed in order for treatment to work and for patients to have a better chance of life. MRI is a very important way to see the shape of brain tumors. However, diagnosing them by hand takes a long time and can be inaccurate depending on who is doing it. We looked at three different versions of the DenseNet201 model to see how well they work at automatically putting brain tumors into four groups: glioma, meningioma, pituitary tumor, and no tumor. The first variant, B-DenseNet201-CA, is trained entirely from scratch to establish a baseline. The second variant, PT-DenseNet201-FB, employs transfer learning with a frozen DenseNet201 base, leveraging pretrained ImageNet features while training only the classification head. The third and most advanced variant, FT-DenseNet201-TL, fine-tunes the last 50 layers of the pretrained DenseNet201 to enhance domain-specific adaptability. A standard MRI dataset was used to train and test all the models. Confusion matrices, accuracy, precision, recall, and F1-scores were used to look at how well they did. The baseline model was only 74.11% accurate, which shows how hard it is to learn from scratch in medical imaging tasks with little data. The frozen transfer learning model significantly improved accuracy to 90.81%, demonstrating the utility of pretrained features. The fine-tuned transfer learning variant further enhanced performance, reaching an accuracy of 94.18% and exhibiting superior class-wise discrimination, particularly for overlapping tumor types. This study shows that partial fine-tuning of a pretrained DenseNet201 is a strong, computationally efficient, and very accurate way to diagnose brain tumors. It also shows how important transfer learning is in medical picture classification. The results could be used in real-life diagnostic tools to help doctors make decisions more quickly and consistently.

Post-Quantum Cryptography: A Fresh Way to Protect Data in Cloud Services

2025-10-04T07:46:21+00:00

This study seeks to apply Post-Quantum Cryptography (PQC) as a newly emerging essentially future secure means of transmitting data within cloud environments. The whole world got up in realization that with the advent of quantum computing, traditional encryption algorithms would be unscrambled. Therefore this proposed research is aimed at developing a quantum resistance security framework to ensure data availability, confidentiality, and integrity for multi-user cloud environments using PQC algorithms, hybrid encryption models, and blockchain-based trust mechanisms.

Social Media Fake News Detection in War Zones: Allied to AI Smooth Detector

2025-10-04T07:48:00+00:00

Fake news and disinformation have emerged as potent tools of modern information warfare, shaping perceptions and influencing outcomes in war zones. Unlike traditional cyber threats that target data or networks, fake news manipulates the human layer of security, exploiting emotions, trust, and social vulnerabilities. This paper investigates SmoothDetector, a probabilistic multimodal AI framework, for detecting fake news on social media during conflict. The system integrates textual analysis, image forensics, contextual credibility scoring, and probabilistic fusion to provide real-time classification and alerts. Experiments on benchmark datasets (FakeNewsNet, LIAR, Twitter15/16) and simulated war-zone streams demonstrate detection accuracies above 90%, resilience to compression and multilingual noise, and real-time detection with latency under 2–7 seconds depending on hardware. The results position SmoothDetector as a viable tool for deployment in command centers, newsrooms, humanitarian agencies, war victims and military information operations. Limitations such as adversarial evasion, false positives, and ethical risks are also discussed, along with a deployment playbook for conflict environments.

AI and Human-AI Collaboration in Software Development: A Comprehensive Analysis of Partnership Paradigms in Modern Development Practices

2025-10-04T10:22:22+00:00

The integration of artificial intelligence into software development has fundamentally transformed traditional programming practices, creating collaborative partnerships between human developers and intelligent systems that redefine the software creation process. This article examines the multifaceted impact of AI-assisted development tools on programming workflows, team dynamics, and skill requirements within contemporary software engineering environments. The article analyzes various AI technologies, including code generation systems, quality assurance tools, and automated testing frameworks, exploring how these innovations complement human capabilities rather than replace them. Through a comprehensive examination of human-AI collaboration patterns, the article identifies emerging skill requirements such as prompt engineering, AI output validation, and collaborative workflow integration that represent essential competencies for modern developers. The article reveals that successful AI implementation requires careful consideration of technical limitations, ethical implications, and quality assurance challenges while maintaining human oversight in strategic decision-making processes. Industry case studies demonstrate measurable improvements in development productivity and code quality when AI tools are properly integrated into existing workflows, though outcomes vary significantly based on implementation strategies and team adaptation approaches. The article concludes that the future of software development lies in sophisticated human-AI partnerships that leverage the complementary strengths of human creativity, contextual understanding, and strategic thinking alongside AI capabilities in pattern recognition, code generation, and routine analysis. This article suggests that organizations must develop comprehensive frameworks for AI integration that preserve essential human skills while maximizing collaborative benefits, ensuring that technological advancement enhances rather than diminishes the creative and analytical contributions that define exceptional software engineering practice.

Certified Domination Polynomials of Generalized Friendship Graphs

2025-10-04T14:04:46+00:00

Let ????=(????,????) be a simple graph of order ????. The certified domination polynomial of ???? is the polynomial ????????????????(????,????)=Σ????????????????(????,????)|????(????)|????=????????????????(????)????????, where ????????????????(????) is the minimum cardinality of certified dominating set of ???? and ????????????????(????,????) is the number of certified dominating sets of ???? of size ????. Let ???? and ????≥3 be any positive integer and ????????,???? be the generalized friendship graph formed by a collection of ???? cycles (all of order ????), meeting at a common vertex. In this article, we study the certified domination polynomials of generalized friendship graphs ????3,????, ????4,???? and ????5,????

RADIO ANTIPODAL CONTRA HARMONIC MEAN GRAPHS

2025-10-04T14:06:32+00:00

In graph theory, graph labeling is a technique that is evolving quickly. A novel graph labeling parameter, radio antipodal contra harmonic mean labeling, has been defined in this study. Radio antipodal contra harmonic mean labeling is also investigated for existence and nonexistence of graphs. Additionally, the procedure for determining the radio antipodal contra harmonic mean number of mean graphs is also displayed.