A Comparative Analysis of Machine Learning Models in Advanced Diabetes Diagnosis

Early identification of diabetes mellitus is critical for efficient therapy and the avoidance of serious implications. Diabetes is a long-term disorder that occurs with higher blood sugar levels which, if not found and treated as soon as it is possible to firm control will end up leading the sufferers into many life-threatening health issues. In this study, we carry out a large dataset containing demographic and lifestyle features as well as extensive clinical factors medical records using machine learning classifiers to predict diabetes. The machine learning classifiers we chose to evaluate and compare were Random Forest (RF), Decision Tree, K-Nearest Neighbors (KNN), Naive Bayes, and Support Vector Machine (SVM), Logistic Regression, Gradient Boosting, AdaBoost, Nearest Centroid, and Multilayer Perceptron. Our experimental results show that RF achieved an F1 score of 99.19% and a best accuracy of 99.18%. The test result proves that this classifier is capable of handling complex information in a dataset as well as providing a reliable prediction under different cases. We also established a user-friendly web application, which takes corresponding inputs from the users and can predict the risk of diabetes. As a practical solution to early diabetes detection and management, it has clinical potential to interface well.