Background and Objective:
Diabetes is one of the most common metabolic diseases worldwide. However, predictive models based on machine learning and artificial intelligence can help predict diabetes more quickly. The present study aimed to predict diabetes using a machine learning approach.Methods: This study was conducted using a systematic review approach and according to the PRISMA guidelines in ۱۴۰۴. Relevant articles were searched for over a ten-year period from PubMed, Web of Science, Scopus, and Google Scholar. The combination of keywords was set using the MeSH system and included the concepts of machine learning algorithm, diabetes, artificial intelligence, prediction, systematic review, and meta-analysis. After the initial search, ۱,۰۶۰ articles were retrieved, and through a multi-stage screening process, ۴۰ articles with desirable quality were finally selected for the final analysis. The findings of these articles formed the basis for a comprehensive evaluation of the applications of artificial intelligence and machine learning in diabetes prediction.Results: Of the ۴۰ eligible articles, a significant variation in the use of machine learning algorithms was observed. Random forest was the most widely used method with ۹ articles (۲۲.۵%). Artificial neural network and support vector machine were next with ۵ articles (۱۲.۵%) each. Logistic regression was used in ۴ articles (۱۰%) and decision tree in ۳ articles (۷.۵%). XGBoost, AdaBoost, combination of decision tree and random forest, and K-Nearest Neighbors algorithms had a similar share with ۲ articles (۵%) each. Gradient Booster, CATBoost, PEM, combination of random forest and logistic regression, and combination of LightGBM with KNN were also observed in ۱ article (۲.۵%) each.Conclusion: Based on the findings, random forest had higher applicability and prediction accuracy in the field of diabetes than other machine learning algorithms. In the next rankings, support vector machine and artificial neural network were identified as the most effective methods, respectively.
