Comparative Study of Machine Learning Algorithms in Predicting Diabetes Onset Using Electronic Health Records

Abstract

The prediction of diabetes onset is critical in enabling early intervention and improving patient outcomes. This study presents a comparative analysis of several machine learning (ML) algorithms applied to Electronic Health Records (EHRs) for predicting diabetes onset. Various ML models, including decision trees, support vector machines (SVM), random forests, and neural networks, were evaluated using clinical data from diverse patient populations. The study assesses the models' performance based on key metrics such as accuracy, precision, recall, F1-score, and ROC-AUC. Results show that while all models demonstrate significant potential for predicting diabetes onset, random forests and neural networks outperform the other algorithms in terms of accuracy and sensitivity. Furthermore, the study highlights the challenges in working with EHR data, such as missing values, feature selection, and data preprocessing. The findings underscore the importance of using diverse ML techniques and improving data quality for better prediction accuracy. This research paves the way for implementing machine learning solutions in clinical settings for early diabetes prediction, ultimately contributing to more personalized healthcare strategies.