A Guided Stochastic Gradient Descent Enhanced Neural Networks for Early Diabetes Readmission Prediction

Diabetes is a prevalent chronic health condition globally, posing significant challenges to ‎healthcare systems and insurance companies due to its associated risks of hospital ‎readmission. Early prediction of readmissions, especially within 30 days, is crucial for ‎directing attention to high-risk patients and optimizing healthcare resources. This study ‎explores the application of machine learning models, including Extreme Gradient Boosting ‎‎(XGBoost), Support Vector Machine (SVM), and Artificial Neural Networks (ANN), to ‎predict diabetic readmissions. A novel approach is proposed, leveraging a guided optimizer ‎for ANN training to enhance classification accuracy and error convergence. Results ‎demonstrate up to a 1.5% improvement in classification accuracy compared to standard ‎methods, highlighting the effectiveness of the guided optimizer in capturing consistent data ‎patterns. By integrating AI-driven predictive analytics, this project aims to improve healthcare ‎efficiency, reduce hospital readmissions, and ultimately enhance patient outcomes in diabetic ‎care. Focusing on type II diabetes, the study addresses dataset challenges in medical ‎bioinformatics and underscores the global significance of mitigating readmissions to alleviate ‎the burden on healthcare systems and improve long-term patient well-being‎.