Traffic Flow Prediction Based on ARIMA Modeling

With the rapid development of urbanization and technology, the number of cars has significantly increased, resulting in increasingly serious urban problems such as traffic congestion, air pollution and frequent accidents. In order to solve these thorny problems, accurate short-term traffic flow prediction is particularly important. This study uses ARIMA model to explore the real-time, stochastic and dynamic characteristics of traffic flow. The data source of this paper is based on the detailed data set provided by spassau. The data set collects one piece of information every 15 minutes for two days, and a total of 200 records are collected. D=0 is obtained by the augmented Dickey Fuller (ADF) test, and it is inferred that the data is stable. Then ARMA (p, q) model is used for fitting and prediction. The test results show that the model performs well in the prediction of 12 steps (3 hours) in advance, the root mean square error (RMSE) is 7.00, and the average absolute percentage error (MAPE) is 11.17%. This method provides a solid technical guarantee for improving the prediction accuracy of urban traffic flow, greatly improving the management level of the city and the living standard of residents.