Despite the greatest efforts of the car industry's engineers and researchers, traffic accidents will continue to occur. In order to better understand the causes of risky traffic occurrences, it would be helpful to design a prediction system that can automatically categorize the severity of injuries sustained in various traffic accidents. Knowing these road and behavioral patterns may help with traffic safety policy making. For policies to be effective, they must be based on rigorous scientific research into the root causes of accidents and the extent of injuries. Multiple machine learning-based injury severity prediction algorithms are available inside the system. In this research, we propose a prediction model for the early identification of traffic accidents using machine learning techniques such as decision trees, Random forests, and support vector machines. We used Kaggle's dataset of traffic accidents. Our proposed approaches to accident prediction have a 98 percent accuracy rate in pinpointing hotspots.