This paper investigates the application of machine learning (ML) models for predicting diabetes using the Pima Indians Diabetes Database, with a focus on enhancing model interpretability through the use of SHapley Additive exPlanations (SHAP). The study evaluates eight ML models, including Adaptive Boosting (AdaBoost), k-Nearest Neighbors (k-NN), Logistic Regression (LR), Multi-layer Perceptron (MLP), Naive Bayes (NB), Random Forest (RF), Support Vector Machine (SVM), and eXtreme Gradient Boosting (XGBoost), utilizing both test/train split and 10-fold cross-validation methods. The RF model demonstrated superior performance, achieving an accuracy of 82% and an F1-score of 0.83 in the test/train split, and an accuracy of 83% and an F1-score of 0.84 in the 10-fold cross-validation. SHAP analysis was employed to identify the most influential predictors, revealing that glucose, BMI, pregnancies, and insulin levels are the key factors in diabetes prediction, aligning with established clinical markers. Additionally, the use of the Synthetic Minority Over-sampling TEchnique (SMOTE) for class balancing and data scaling contributes to robust model performance. The study emphasizes the necessity for interpretable ML in healthcare, proposing SHAP as a valuable tool for bridging predictive accuracy and clinical transparency in diabetes diagnostics.