A Comprehensive Analysis and Comparative Insights for Developing Diabetes Prediction Models

This research aims to contribute to the on-going research to develop robust and pragmatic models for predicting the condition of diabetes in individuals. While it is acknowledged that achieving 100% accuracy in predicting diabetes is challenging due to the complex and numerous parameters involved, this study focuses on designing and training models that can significantly improve the performance of healthcare sectors in dealing with diabetes. To address this objective, the following actions were undertaken: (i) A comprehensive review of over 30 research papers related to the topic, (ii) Thorough comparison and evaluation of insights derived from the literature, (iii) Building, experimenting and applying different machine learning techniques and deep learning approach on the ‘130 US Clinical Hospitals dataset’ and ‘PIMA’ dataset, (iv) Comparison of the developed models with the ones suggested and mentioned in the existing research papers, (v) Presenting fresh machine learning and deep learning models to use to train diabetes models. While initially intended as a comparative analysis, this research goes beyond by providing strong conclusions based on the conducted investigations.