USE OF MACHINE LEARNING ALGORITHMS IN DIABETES DETECTION

The utilization of machine learning algorithms in the medical domain has gained significant traction, particularly in predicting disease progression, identifying complications, and aiding in clinical decision-making. This study investigates the application of various machine learning techniques to predict the presence of diabetes, leveraging an extensive dataset derived from patient medical records. The dataset encompasses crucial medical indicators that contribute to the classification process. Preliminary data processing involved addressing missing values, normalizing the data, and partitioning it into training and testing subsets. The research evaluates the efficacy of five distinct algorithms: logistic regression, decision tree, random forest, support vector machine (SVM), and k-nearest neighbors (KNN), to determine the most effective model for diabetes prediction. Performance metrics were employed to assess each algorithm’s predictive accuracy and reliability. The results indicate that KNN demonstrates superior performance, effectively distinguishing between positive and negative cases of diabetes. In contrast, logistic regression exhibited commendable overall efficacy, whereas the decision tree algorithm revealed limitations in reliability. Both random forest and SVM models produced satisfactory results; however, they presented specific constraints in particular contexts. This study underscores the potential of machine learning algorithms in diabetes diagnostics, highlighting the critical importance of algorithm selection tailored to clinical scenarios. Future research should aim to enhance data integrity, refine algorithmic approaches, and optimize hyperparameters, thereby augmenting predictive accuracy. Such advancements are poised to significantly improve diabetes diagnostics and facilitate timely patient identification