Comparison of Classification of Different Machine learning Algorithms in the Diagnosis and Detect of Diabetes

Abstract

Diabetes, caused by a rise in blood glucose levels, can be detected using a variety of instruments that analyze blood samples. Heart attacks and kidney failure are among the serious complications that can arise from untreated diabetes. Consequently, the field of detecting and evaluating gestational diabetes requires more robust research and better learning models. The information system for detecting diabetes in this study is based on machine learning (ML) algorithms. In the study, various machine learning techniques are discussed, including Decision Trees (DT), Random Forest (RF), Logistic Regression (LR), Extreme Gradient Boosting (XGBoost), and K-Nearest Neighbors (K-NN). The data was collected from the Iraqi society, mainly from the laboratory of Medical City Hospital and the Specialises Centre for Endocrinology and Diabetes-Al-Kindy Teaching Hospital. On the basis of the Recursive Feature Elimination approach, research has been done to enhance the prediction index. The performances of all five algorithms are evaluated on the various measures like the Precision, Accuracy, F-Measure, Recall, Cohen Kappa, and AUC. Accuracy is measured over correctly and incorrectly classified instances. The Results obtained show XGBoost outperforms with the highest accuracy of 98% comparatively other algorithms. This study's findings can be inform a program for screening potential diabetes patients.