Diabetes, a common metabolic disorder, poses serious health risks and can lead to a variety of complications if not treated. Early detection is critical for mitigating these risks, making accurate prediction methods essential. In this study, we look at how various machine learning classifiers (KNN, Logistic Regression, Random Forest, SVM with Linear, RBF, and Polynomial Kernels, and Gradient Boosting) perform in categorizing diabetic patients. We evaluate these classifiers’ performance using metrics like accuracy using data from the UCI ML repository. Furthermore, we use feature engineering techniques to preprocess and clean the data, improving the quality of input features and resulting in better model performance. Through our comparative analysis, we hope to identify optimal approaches for diabetes prediction, contributing to improved diagnosis.