Investigating Role of Supervised Machine Learning Approach in Classification of Diabetic Patient

Abstract

Objectives: Healthcare analytics requires classifying diabetic patient datasets for quicker diagnosis and personalized treatment. This study used SVM, Decision Trees, KNN, ANN, and Logistic Regression to predict type 1 and 2 diabetes. Our detailed performance research shows these algorithms' utility in handling diabetic patient data's complexity. Methods: We compare SVM, Decision Trees, KNN, ANN, and Logistic Regression for diabetes patient dataset classification. While each approach has pros and cons, ANN and logistic regression are promising clinical possibilities. Diagnoses, proactive therapies, and diabetes patient outcomes increase with these breakthroughs. Results: SVM has 84.3% accuracy in type 1 and type 2 diabetics. SVM recognized complicated dataset patterns with great accuracy and recall. Decision trees were more interpretable and could record diverse choice limits, with accuracy rates of 86.15% for both types. With 90.6% accuracy, KNN predicted type 1 diabetes well. KNN was ideal for complex datasets because to its greater accuracy and recall using data point similarities. ANN and Logistic Regression had the highest accuracy for type 1 and type 2 diabetes patients at 96.1% and 97%, respectively. Novelty: Layered ANN and logistic regression identified complicated dataset relationships with accuracy, recall, and F1 scores exceeding 0.94. ANN with logistic regression may change diabetes patient classification with unequaled prediction power and accuracy.