Intuitive Feature extraction techniques with IDDN Model for Women’s Diabetic Prediction

Abstract

Diabetes prediction for women, especially around pregnancy, is critical due to the elevated
risk of gestational diabetes and its long-term health effects. This study presents an advanced
approach utilizing the Interpolative Deep Dense Layer Design (IDDLD) algorithm to enhance
prediction accuracy and interpretability for complex, multi-phase datasets related to women's
health. The IDDLD algorithm improves traditional deep learning frameworks by integrating
interpolative techniques within dense neural networks, addressing the dynamic changes in
health status before and after pregnancy. By incorporating new features such as indicators for
heart disease (based on elevated glucose and blood pressure), kidney disease (high insulin
and skin thickness), eye disease (BMI and diabetes pedigree function), and gastrointestinal
diabetes (age and glucose levels), the model captures critical health markers associated with
diabetes. Evaluations of the IDDLD model using pre-pregnancy and post-pregnancy health
records demonstrated significant improvements in predictive accuracy, with higher precision
and recall rates compared to traditional models. The algorithm’s ability to adapt to temporal
variations and physiological changes offers enhanced interpretability and actionable insights,
contributing to more personalized and effective healthcare strategies. This research highlights
the potential of the IDDLD model to advance diabetes prediction methodologies, supporting
better health outcomes for women across different life stages.