Diabetes Mellitus Prediction and Severity Calculation Using Stacked Ensemble Method

Millions of individuals throughout the world suffer with diabetes mellitus, a chronic condition that can be effectively managed with early detection and precise prognosis. In this work, a machine learning method for predicting diabetes utilizes stacking ensemble algorithms with hyperparameter tuning. The proposed method uses three fundamental classifiers such as Random Forest, logistic regression, gradient boosting and tunes certain hyperparameters to improve the performance of each algorithm. Another classifier, called support vector machine, with hyperparameter tweaking is used as a meta-classifier to aggregate the predictions of the basic classifiers. It has been demonstrated that the stacking ensemble technique, especially when combined with hyperparameter tuning, enhances the prediction performance of machine learning models. The proposed technique is evaluated using two datasets, namely PIMA Indian Diabetes dataset, and dataset 2. Hierarchical clustering is used to forecast the disease's severity level. Severity levels are assigned based on the model's clustering results. The findings show that the suggested method accurately predicts diabetes and performs better than the individual base classifiers. To evaluate the severity level, a model created using a machine learning classification technique is chosen for further development. The model is finally included with the web application. The efficiency of the model is evaluated using a variety of performance criteria for evaluation such as accuracy, precision, recall, and F1-score. The proposed stacked ensemble method results in the accuracy of 97% which shows the improvement in the state-of-the-art methods.