Publication

Abstract : Diabetes is a long-term metabolic disorder and a growing global health concern, and hence, early, and proper diagnosis is critical to prevent serious complications. In this paper, the Pima Indian Diabetes dataset is utilized to create machine learning algorithms that have a high degree of accuracy in diabetic prediction. Following extensive pre-processing for data quality, variety of ensemble machine learning models are used for example Random Forests, Gradient Boosting, Stacking, and Voting Classifier. The model's performance was evaluated using the F1-score, recall, accuracy, & precision metrics. Among the mentioned models, the Stacking model performed the most optimally at a testing accuracy of 77.27 %. Feature importance calculation identified “Glucose” and “BMI” as the highly significant predictors. These findings show the promise of machine learning to improve early diabetes diagnosis and facilitate better clinical decision-making. Future work will explore better generalization by employing more advanced algorithms and more diverse data sets.