Performance Analysis of Machine Learning Algorithms for Diabetes Diagnosis Prediction

Diabetes is a chronic condition characterized by increasing blood sugar levels above average. In the long term, diabetes can have damaging effects on the kidneys, heart, nerves, eyes, and other complications. Early detection in the treatment of diabetes can prevent and reduce the risk of further disease complications. The field of diabetes prediction is experiencing significant growth in research endeavors. Early diabetes prediction will lead to improved treatment. In this study, we propose predicting the diagnosis of diabetes using machine learning algorithms. The dataset utilized in this research is the Pima Indians Diabetes Dataset (PIDD), sourced from the UCI machine learning repository. Seven machine learning classification algorithms are applied: Support Vector Machine (SVM), Naïve Bayes, K-Nearest Neighbor (KNN), Logistic Regression, Random Forest, Decision Tree, and Gradient Boosting. Various measures, including Accuracy, Precision, F-Measure, Recall, and Receiver Operating Characteristic (ROC), evaluate the performance of the seven machine learning algorithms. The performance results show that the KNN algorithm obtains the highest accuracy of 80.51% and Precision of 78%. The Random Forest algorithm gets the highest score on the Recall at 75%, F-Measure at 72%, and ROC at 78.18%.