Diabetes is a prevalent and enduring condition impacting millions of individuals worldwide this is why early detection is essential for efficient management and intervention. The objective of this article is to build a trustworthy diabetes prediction model using the RF-SVM algorithm and employing ensemble techniques, specifically the stacking technique. The Pima Indians dataset, renowned for its comprehensive clinical and demographic features, is utilized for this study. Our proposed RF-SVM ensemble model trains on a subset of the dataset using stratified cross-validation to increase robustness and generalizability. With an accuracy of 86%, the proposed model successfully predicts diabetes, demonstrating its value in early diagnosis and timely treatment. Feature importance analysis can help us better understand the variables that affect how diabetes develops. This study demonstrates the utility of the RF-SVM ensemble model with the stacking technique for diabetes prediction. The developed approach is effective in identifying patients who are at risk, which improves patient outcomes. Future research initiatives may include merging more datasets and researching advanced machine learning approaches to increase prediction accuracy and increase the model’s utility.