Advances in diabetes prediction: a systematic literature review of Artificial Intelligence based methods

Diabetes mellitus (DM), a common glycemic condition that causes substantial challenges to public health. The growths of Artificial Intelligence (AI) have created notable change in predicting DM, offering novel possibilities to lower its effect. This comprehensive review examined 25 articles concerning machine learning (ML) uses for DM prediction, emphasizing datasets, models, and evaluation techniques. Several datasets, including the Pima Indians Diabetes Database (PIDD), the National Health and Nutrition Examination Survey (NHANES), and REPLACE-BG, have been analyzed, highlighting their typical features and related issues, such as unbalanced data. This study evaluates the efficiency of various ML algorithms, including Support Vector Machines (SVM), Logistic Regression, XGBoost, and Convolutional Neural Networks (CNN), in predicting DM across several datasets. A few validation techniques are discussed, including k-fold cross-validation, and evaluation metrics including area under the curve, accuracy, sensitivity, and specificity. The result shows the importance of ML in handling the issues associated with DM prediction, and the need of maintaining models therapeutic relevance. With the ultimate goal of reducing the prevalence of this common disorder, this review helps current capability to use AI methods for better DM prediction.