AN ENHANCED ASTUTE SYSTEM FOR PERSONALISED DIABETES DIAGNOSIS

Insulin resistance is the root cause of diabetes, a worldwide epidemic of chronic illness. Due to the lack of a cure for diabetes, the only method to lessen the threat of complications from the condition is via early diagnosis and treatment. Research towards the early diagnosis of diabetes using machine learning methods has been extensive. Improving prediction accuracy, however, is notoriously challenging due to the presence of missing values, irrelevant information, and uneven class distribution in the dataset. To better categorize the Pima Indians Diabetes Dataset (PIDD), we provide a Tree-Based machine learning approach in this study. To improve predictions, a Mutual Information (MI)-based feature selection approach is used to filter out irrelevant data. Finally, the Adaptive Boosting (AB) algorithm is used to boost the efficiency of Tree-Based algorithms. The Extra Tree (ET) technique used as the basis estimator of the AdaBoost classifier has the best accuracy (90.5% accuracy) when tested against experimental data. As a result, our suggested Tree-Based ML model may help doctors with diabetes diagnosis.