Diabetes mellitus (DM) affects the hormone insulin, which causes improper glucose metabolism and raises the body’s blood sugar levels. With 4.2 million fatalities in 2019, DM is one of the top 10 global causes of mortality. Early detection of DM will aid in its treatment and avert complications. There must be a quick and simple technique to diagnose it. Such diseases can be managed, and human lives can be saved with early diagnosis. Smart prediction techniques like machine learning (ML) have produced encouraging outcomes in predictive classifications. There has been a lot of interest in ML-based decision-support platforms for the prediction of chronic illnesses to provide improved diagnosis and prognosis help to medical professionals and the general population. By building predictive models using diagnostic medical datasets gathered from DM patients, ML algorithms efficiently extract knowledge that helps predict diabetic individuals. The association between DM and a healthy lifestyle is used in the model. In this study, the NHANES (National Health and Nutrition Examination Survey) data set is utilized, along with five ML methods such as Artificial Neural Networks (ANN), CATBoost, XGBoost, XGBoost-histogram, and Light GBM to predict DM. The results of the experiment demonstrate that the XGB-h model outperformed other ML methods regarding area under the receiver operating characteristic curve (AUC-ROC) and accuracy. The most effective XGB-h framework can be used in a mobile app and a website to rapidly forecast DM. Real-time prediction using details delivered by the model at runtime can be developed as a whole bundle as a product. Clinicians can quickly determine who is likely to get diabetes using the proposed strategy, which will facilitate prompt intervention and caring.