Diabetes is a well-known chronic disease that includes a range of metabolic disorders characterized by persistently elevated blood sugar levels over an extended period of time. Early and precise prediction of diabetes is essential to reduce risk factors and minimize potential complications associated with the disease. However, there are significant challenges in creating reliable predictive models due to factors such as limited labeling data, the presence of outliers, and the absence of information in diabetes-related datasets. To address these barriers, this paper proposes a comprehensive framework aimed at improving diabetes prediction through data preprocessing and machine learning techniques. The framework combines methods for dealing with missing values, data standardization, and feature visualization to extract meaningful insights. In addition, various machine learning classifiers - including support vector machine (SVM), decision tree, logistic regression, and naive Baye - are implemented to improve prediction accuracy and support early diagnosis of diabetes. Among these models, SVM shows better comprehensive performance.
