Objective To quantitatively analyze the development trend of traffic accidents and provide decision support for precise prevention and control of urban traffic safety management. Methods By collecting data related to highway traffic accidents in a certain city from 2010 to 2020,analyzing the correlation between data attributes and traffic accidents,a prediction model for highway traffic accidents in a certain city was established based on the ARIMA method.The reliability of the ARIMA model was verified by comparing the fit of the exponential regression curve model and the grey optimization GOM model. Results The prediction accuracy of the ARIMA (4,1,3) model established by fitting traffic accident time series data based on discreteness and randomness was over 94%,and the prediction performance was significantly better than the exponential regression curve model and the grey optimization GOM model.The model had good reliability and applicability. Conclusion The traffic accident prediction model established based on ARIMA model has certain guiding significance for the refined management of highway traffic safety in a certain city.