A Driver Injury Prediction Model based on Genetic Algorithm and BP Neural Network

In order to improve the survival rate of the injured in the accident, many vehicles are now equipped with automatic crash notification system (ACNS) in vehicle. As the core of the system, the driver injury prediction model can predict the driver's injury category in time and send the injury situation to the emergency medical institution. The medical institution arranges the optimal rescue team and hospital according to the injury situation obtained by the algorithm, which greatly reduces the economic loss and mortality caused by the accident. This paper mainly studies the severity of driver injury. Using the annual data from 2019 National Highway Traffic Safety Administration (NHTSA) Fatality Analysis Reporting System (FARS), from which 10 variables such as driver injury, driver age and height were extracted, and the correlation between each independent variable and dependent variable was analyzed to increase the reliability and prediction accuracy of the network model adopted in this paper. In this paper, the combination of genetic algorithm and BP neural network was used to build a driver injury prediction model with machine learning. Compared with support vector machines (SVM), long short-term memory (LSTM), traditional BP neural network and logistic linear model, the accuracy was improved by 11.78%, 6.54%, 7.08% and 13.78% respectively. The research results can be used to improve the algorithm and performance of the enterprise call center in the advanced vehicle collision automatic call system, and finally can effectively improve the efficiency of accident rescue.