Machine Learning-Based Cellular Traffic Prediction Using Data Reduction Techniques

Abstract

For modern networks to maximise Quality of Service (QoS), precise cellular traffic prediction is crucial, particularly given the increasing need for real-time applications. In order to increase predictive performance, this study introduces an upgraded Adaptive Machine Learning-based Cellular Traffic Prediction (AML-CTP) framework that incorporates cutting-edge machine learning algorithms such as XGBoost, CatBoost, and Voting Regression with parameter adjustment. The suggested system makes use of density-based clustering techniques like DBSCAN to concentrate on high-similarity data clusters, PCA for dimensionality reduction, and robust preprocessing approaches like Min-Max Scaling. These techniques minimise computing complexity while guaranteeing effective model training. The Flask framework is used in the system's implementation to improve accessibility and real-time deployment, enabling smooth user interaction for data uploading and prediction generation. With the greatest R2 score of 98%, experimental findings illustrate how successful XGBoost is at adapting to changing traffic patterns, allocating resources optimally, and improving overall quality of service (QoS) in cellular networks.