Road Accident Prediction using Machine Learning

Abstract

Between 3% and 5% of the GDP is lost each year due to traffic accidents. To be more specific, India was responsible for more than 6% of all traffic mishaps worldwide, even though the country only has around 1% of the world's road vehicles. Teens and young adults were impacted by almost 70% of the occurrences. Researchers found that the death-to-injury ratio would rise in the event of vehicle accidents. It is considered an urgent requirement to use modern techniques for traffic planning and management. Reduced car collisions are an inevitable outcome of regulations and initiatives based on the recognition of road hazards. Building a presumptive model using current data and potential future risks can be helpful.
We proposed a model for early crash detection predictions using machine learning methods including SVM, Random Forest, and decision trees. The aforementioned research led to the conclusion that variables such vehicle type, device code, equipment position, location latitude and longitude, speed, date and time, etc. are crucial in road accident investigations. We made use of the collision data provided by Kaggle. We attained a 98 percent accuracy rate for incident forecasting using the methods we suggested for finding the location with the greatest accident frequency. The investigation and identification of the technique and causes of the accidents would be aided by this estimate.