Prediction of Several Machine Learning Classification Models to Predict Type 2 Diabetes Diagnosis in Females

Abstract

Medical and biological sciences are among the many fields greatly impacted by machine learning (ML). Diabetes is a chronic disease characterized by unusually high blood sugar levels and insulin use. The analysis of diabetic patients and the diagnosis of the disease using various ML approaches is the main subject of this research. In this study, ML models were estimated using the PIMA dataset, which consists of female diabetic patients who were at least 21 years old. The relevant dataset was originally provided by the National Institute of Diabetes and Digestive and Kidney Diseases. Various diagnostic measurements are included in the dataset. Model performance was evaluated using an exploratory dataset and the 5-fold cross-validation method. The purpose of the dataset is to predict whether a patient has diabetes using diagnostic ML algorithms. In this study, the performance of classifiers such as Naive Bayes (NB), Stochastic Gradient Boosting (SGB), Extreme Gradient Boosting (XGBoost), and Logistic Regression (LR) algorithms were compared. According to the results of the performance measurements for diabetes prediction, the Stochastic Gradient Boosting (SGB) model outperformed the other ML algorithms in terms of Accuracy, Balanced Accuracy, F1-Score, Sensitivity, Specificity, Youden index, Positive Predictive Value (PPV), and Negative Predictive Value (NPV).