As we can see, diabetes, a chronic disease, is increasing daily, leading to weight gain, blindness, abnormal cholesterol levels, etc. The most archetypal indications of this disease are kidney failure, blindness, and the more rapid heart failure are increasingly prevalent among individuals with diabetes. These trends not only pose significant challenges to healthcare systems worldwide but also reveals how critical it is to develop and apply effective prevention and treatment methods. Our study highlights a recent article published by the Lancet, stating that by 2050, diabetes is expected to affect 1.31 billion people. This number is significantly higher than the 529 million reported in 2021. This shows that the issue of increasing number of people suffering from diabetes as a global health problem is becoming more serious. The purpose of this study is to prevent diseases that cause death. Therefore, nowadays AI technology is used to detect such diseases. We used several different machine and AI techniques to predict such outliers in this research paper. The technique we used compares different algorithms to locate the highest quality one for predicting diabetes. Another method used is anomaly detection which is employed to detect unusual signs that might indicate diabetes and the third is the stacking classifier which consists of multiple algorithms for final predictions. After applying these techniques, we found that the stacking classifier gives the best performance, achieving 84% accuracy which improves the robustness of the model. Our research on anomaly showed that the dbscan method is the best for detecting outliers. These machine learning approaches demonstrate how they increase performance and reliability of accuracy.