Prediction of Injury Severity and Vehicle Deformity using Machine Learning Approaches on Heterogeneous Real Time Accident Datasets

Abstract

In recent years, traffic accidents have become a frequent cause of death. According to World Health Organization WHO the number of deaths on the worlds roads remains unacceptably high with 1.35 million people dying each year. Public security in road accidents is the worlds most important goal. Finding the cause of death in an accident has always been a motive for transportation. newlineThe research work proposes road accident prediction for National Highway Traffic Safety Administration NHTSA Fatality Analysis Reporting System FARS dataset using optimized newlineeXtreme Gradient Boosting XGBoost machine learning model. The research work for road accident prediction includes prediction of injury severity of person for different types of crash like angles, collision between motor vehicles and prediction of extent of damage in vehicle in crash. newlineHuman injury in a vehicle crash is a critical subject of analysis. The injury severity of a person in crashes helps the transportation agency to determine crash conditions. This will in turn helps road safety manager or engineer to implement the counter measures and to improve and enhance the level of safety at the roadside. Prediction of injury severity of person in road traffic accident is considered as a important and crucial analysis for driving decisions under the dangerous situations. For the prediction of injury severity, four different hybrid machine learning approaches are proposed which address the issue. newline The proposed systems for injury severity prediction and extent of damage in vehicle prediction includes following steps, selecting the dataset, data pre processing, features selection to find the important attributes, classification and performance evaluation. For prediction and evaluation, we use NHTSA FARS dataset in first step. NHTSA has an National Center for Statistics and Analysis NCSA data collection system that provides accident related data which includes data on fatal accidents in 50 states in the United States.