Despite advancements in autonomous driving technology, the frequent occurrence of traffic accidents, particularly in complex traffic environments, indicates that traffic safety remains a significant concern and that substantial safety risks persist in autonomous driving. One of the most critical risks to consider at this stage is driving safety. Managing security risks requires the use of accurate risk prediction models and efficient decision-support systems. In this context, driving risk assessment and predictive decision- support systems play a crucial role in improving the safety of autonomous vehicles. They commonly employed in research include statistical regression models and machine learning algorithms. These methodologies, particularly those using machine learning algorithms, have significantly enhanced the accuracy of risk predicts and have been widely adopted by researchers in recent years. The data used in these studies can be obtained from environmental traffic conditions, environmental parameters, in-vehicle devices, cameras, sensors, or naturalistic driving datasets. This study comprehensively reviews the current literature on Machine Learning-Based Risk Prediction and Decision Support Systems in Autonomous Driving using mentioned datasets. Based on the literature findings, the study evaluates the limitations of existing research, future research directions, and the reliability of ma-chine learning models. This review demonstrates how machine learning techniques can be effectively applied to enhance driving safety and reduce accident rates.