<![CDATA[Diabetes mellitus is a chronic health condition that continues to increase worldwide and may lead to various severe complications when early risk identification is not performed. This study addresses the limited implementation of machine learning models in practical web-based applications that can assist users in analyzing diabetes risk based on health-related parameters. Therefore, this research aims to develop DiAnalyze, a web-based diabetes risk prediction platform utilizing the Random Forest algorithm. The study used the Diabetes Prediction Dataset obtained from Kaggle, initially containing 100,000 records with eight input features and one target variable. After data preprocessing and class balancing using undersampling, 18,482 records were retained and divided into training, validation, and testing sets using a 64:16:20 ratio. The Random Forest model achieved an accuracy of 90.70%, precision of 89.09%, recall of 90.97%, F1-score of 90.02%, and ROC-AUC of 97.61%. The trained model was successfully integrated into the DiAnalyze web-based platform, enabling users to obtain diabetes risk classifications and corresponding risk probabilities based on the provided health parameters. DiAnalyze is intended as a supporting tool for diabetes risk assessment and does not replace professional medical diagnosis.]]>
Copyrights © 2026
