Diabetes Mellitus is a chronic disease and one of the non-communicable diseases whose growth is very fast. This study aims to explore and analyze the early detection and prediction system of risk factors for type 2 diabetes mellitus which utilizes machine learning methods. This type of research is a scoping review to accumulate and synthesize the results of previous studies on the early detection of risk factors and the prediction system of Diabetes Mellitus type 2 using machine learning methods. The inclusion criteria are articles in English or Indonesian, journals published in the 2017-2021 range, full text, and not systematic reviews. Article searches are 4 databases, namely Google Scholar, Pubmed, International Journal of Public Health Science/Hindawi, and IEEE Xplore. The results obtained as many as 2,941 articles, using the PRISMA method. The remaining 15 studies were maintained and met the criteria for qualitative analysis. The articles used machine learning methods in the creation of early detection models and prediction systems. Some articles use the merging of two methods (statistical and machine learning). The machine learning techniques mostly use supervised, unsupervised, and deep learning techniques. For the algorithms used, the majority of researchers used more than one algorithm such as algorithm support vector machine (SVM), random forest (RF), Decision Tree (DT), LASSO, and others, to compare the best accuracy of each algorithm. Risk factors associated with Diabetes Mellitus type 2 incidence are age, gender, obesity, family history of the disease, lack of physical activity, genetics, environment, smoking, blood pressure, and diet.