<![CDATA[Diabetes mellitus (DM) is a chronic condition requiring lifelong management due to inadequate insulin secretion or inefficacy of insulin. Its global prevalence has led to extensive research focusing on diagnosis, prevention, and treatment. The developments in artificial intelligence (AI) have improved diabetes management and prediction. This paper provides a comprehensive review of the contributions of machine learning (ML) algorithms in predicting and classifying diabetes. The review examines research on artificial intelligence techniques used to predict diabetes over the past six years, intending to identify the latest innovations and trends in this field. This time frame reflects recent methodological advances and new applications that exemplify the current state of artificial intelligence in diabetes prediction. It covers dataset selection, preprocessing, AI algorithms application, and evaluation methodologies. The results of this review show that the most predominant methods used in diabetes prediction are Random Forest, Logistic Regression, Decision Trees, Support Vector Machine, Naïve Bayes, and K-Nearest Neighbors, each with distinct advantages and limitations. The review also shows through its examination that the highest accuracy provided by the hybrid approach was 99.4%, the ensemble approach (ada boost) was 98.8%, deep learning (DNN) was 98.04%, and traditional machine learning (decision tree_ ID3) was 99%. Most studies conducted for diabetes prediction trained the models on specific datasets, which makes their generalizability to diverse populations and healthcare settings limited. The future directions must address ensuring the robustness and generalizability of predictive models through comprehensive external validation across various populations, settings, and geographic areas.]]>
