Accident Prediction and Analysis using Machine Learning models

Road safety is a major concern due to the rapidly growing number of automobiles. Every year, 1.2 million individuals die in road accidents According to statistics, between 2017 and 2020, traffic accidents in Tamil Nadu resulted in the deaths of almost 22,000 individuals. According to the “Accidental Deaths & Suicides in India-2021” report released by the National Crime Records Bureau (NCRB) shows Tamil Nadu at second in the number of fatalities on roads caused by negligence. The State came in second to Uttar Pradesh, which recorded 18,972 fatalities in 18,228 accidents, with 15,384 fatalities in 14,747 occurrences. The proposed work is used to predict the severity of an accident at a place and time using few machine learning algorithms namely decision tree, random forest, logistic regression, and decision tree hyper parameter tuning. Various parameters are found, such as the speed limit, age, weather, vehicle type, light conditions, and day of the week had been used to train the model. Upon inferring the results, we found that random forest algorithm showed the highest accuracy of these algorithms, with a score of 88.89%, followed by logistic regression (86.23%), decision tree hyper parameter tuning (85.74%), and decision tree (66.67%), respectively. The major challenges faced in this project was collecting proper samples for training the models and modifying it as per the needs of the proposed model. This model will be crucial for traffic planning and management and will aid in the future reduction of many traffic accidents.