This study proposes a machine learning (ML) technique for predicting diabetes through a mobile app, addressing the critical need for early detection and prediction in Saudi Arabia. The research utilizes both the Pima Indians dataset and a private diabetes dataset, employing a semisupervised model combined with strong gradient boosting. The Synthetic Minority Over-sampling Technique (SMOTE) is used to handle imbalanced classes, and ten ML classification techniques are evaluated to determine the most accurate algorithm. The XGBoost algorithm, with SMOTE, achieved impressive performance, achieving an accuracy of 97.4% and an F1 coefficient of 0.95 on the private dataset, and an accuracy of 83.1% and an F1 coefficient of 0.76 on the combined datasets. An explainable AI technique using SHAP methods is implemented to understand the model's predictions, and a mobile app is developed for instant diabetes prediction based on user-entered features. The study contributes novel insights and techniques to the field of ML-based diabetic prediction, potentially aiding in early detection and management of diabetes in Saudi Arabia.This study proposes a machine learning (ML) technique for predicting diabetes through a mobile app, addressing the critical need for early detection and prediction in Saudi Arabia. The research utilizes both the Pima Indians dataset and a private diabetes dataset, employing a semisupervised model combined with strong gradient boosting. The Synthetic Minority Over-sampling Technique (SMOTE) is used to handle imbalanced classes, and ten ML classification techniques are evaluated to determine the most accurate algorithm. The XGBoost algorithm, with SMOTE, achieved impressive performance, achieving an accuracy of 97.4% and an F1 coefficient of 0.95 on the private dataset, and an accuracy of 83.1% and an F1 coefficient of 0.76 on the combined datasets. An explainable AI technique using SHAP methods is implemented to understand the model's predictions, and a mobile app is developed for instant diabetes prediction based on user-entered features. The study contributes novel insights and techniques to the field of ML-based diabetic prediction, potentially aiding in early detection and management of diabetes in Saudi Arabia.