Comparison of Machine Learning Classification Methods for Early Detection of Diabetes

Diabetes is a common disease and is a significant issue in developing and developed countries.Machine learning, a branch of artificial intelligence, provides various algorithms that can efficiently process and analyze medical data to make accurate predictions.This study investigates several machine learning classification approaches for diabetes prediction from medical data sets.This study uses the Pima Indians Diabetes Database for evaluation purposes.These research uses algorithms include Logistic Regression (LR), Decision Tree (DT), Random Forest (RF), Support Vector Machine (SVM), Naive Bayes (NB), Artificial Neural Networks (ANN), K-Nearest Neighbor (KNN), XGBoost (XGB), Linear Discriminant Analysis (LDA), and Quadratic Discriminant Analysis (QDA).This research has, as the ultimate goal, a comparison and assessment of the performance of these algorithms on diabetes prediction.Among all the above-mentioned algorithms, LR is the best-performing model for precision (72.9%) as well as for accuracy (77.73%) and F1score (65.04%).The KNN model was best illustrated by recall (63.42%).The LDA model produced the maximum Area Under the Curve (AUC) value as 83.9%.