Аннотация
Diabetes mellitus poses a significant global health challenge, with increasing prevalence, particularly in low socioeconomic regions. Accurate and early diagnosis is crucial to prevent the severe long-term complications associated with diabetes. This study conducts a comprehensive comparison of six prominent machine learning algorithms-K-Nearest Neighbors (K-NN), Naive Bayes, Support Vector Machine (SVM), Decision Trees, Random Forest, and Logistic Regression-in predicting diabetes using a dataset of 768 individuals with diverse diabetic indicators from Kaggle. Each algorithm is rigorously evaluated based on precision, recall, and F1-score to determine the most effective method for diabetes diagnosis. The results indicate that Logistic Regression outperforms the other algorithms, achieving an accuracy of 81%. This superior performance is attributed to Logistic Regression's ability to effectively delineate linear separations, which is crucial for distinguishing between diabetic and non-diabetic individuals. The study underscores the importance of feature selection and model tuning in enhancing predictive performance. The findings suggest that integrating Logistic Regression into clinical settings can significantly improve the accuracy and timeliness of diabetes diagnosis, potentially leading to better patient outcomes and reduced healthcare costs.