Medium-Term Forecast of Metro based on Big Traffic Data

Abstract

With the rise of industrialization and automation, the demand for transportation has been continuously increasing. The demands for quick and effective travel can no longer be met by conventional transportation techniques. In the field of transportation, urban train travel is significant. At the same time, passenger flow prediction has become increasingly important. In addition to short-term forecasts, medium-term forecasts of daily intervals of passenger flow are also essential. This research employs the ARIMA model to forecast medium-term passenger flow, utilizing subway passenger flow data collected in Chengdu City from April 29th to September 28th, 2019. Based on the data and methodology, this study investigates the feasibility of using the ARIMA model for medium-term subway passenger flow prediction. The research findings indicate that the model exhibits strong seasonality, with an ADF unit root test p-value less than 0.05 and an LB statistic test p-value less than 0.05. The Mean Percentage Error is only -1.03%. Furthermore, there is a strong concordance between the fluctuation patterns observed in the actual and predicted values of the model, which closely align with the trends observed in the training dataset. Through this research, it is concluded that the ARIMA model performs well in medium-term subway passenger flow prediction.