Diabetes, an inescapable and constant condition, presents critical worldwide difficulties, requiring early identification and viable administration to further develop wellbeing results. As of late, AI calculations applied to wellbeing records have shown guarantee in foreseeing diabetes risk. Advancing the hyperparameters of these AI models is essential, as it fundamentally influences their exhibition. This exploration study investigates the capability of AI strategies and hyperparameter tuning for diabetes prediction. In particular, the framework uses the Pima Indians Diabetes dataset and applies the whale improvement calculation to tweak the hyperparameters of the XGBoost calculation. The essential objective is to improve the exactness and productivity of diabetes prediction, adding to early discovery and counteraction. The review’s outcomes show the proposed approach’s cutthroat exhibition contrasted with customary techniques, validating its adequacy in this field.