Long-Short-Term Memory Networks for Diabetes Disease Prediction Using Artificial Intelligence

Abstract

This research investigates the application of AI-driven Long Short-Term Memory (LSTM) networks for the prediction of diabetic illness. LSTM models are well-suited for analyzing the temporal nature of diabetes-related data, enabling accurate forecasting of disease progression and associated complications. The primary objective is to develop LSTM architectures capable of capturing complex patterns in longitudinal patient data such as blood glucose levels, medication adherence, and lifestyle behaviors and to evaluate their predictive performance against traditional statistical and machine learning methods. By leveraging artificial intelligence, this study aims to deliver timely insights into the trajectory of diabetes, thereby empowering both patients and healthcare providers to make more informed decisions. These insights can facilitate the optimization of treatment strategies and enhance preventive care to reduce diabetes-related complications. The dataset utilized in this study is sourced from Data World, encompassing various aspects of diabetes prediction, classification, and healthcare outcomes. One component of the dataset indicates an equal distribution of diabetic and non-diabetic patients (5 out of 10 each). Other dataset segments highlight high BMI and elevated cholesterol levels as significant predictors of diabetes in individuals aged 60–70. Additionally, factors such as smoking, heart disease, and BMI are found to influence diabetes risk across both male and female demographic groups in separate database subsets.