Travel time characteristics of subway passenger flow and short-time passenger flow forecast

This thesis is based on large-scale travel data, first of all, the metro passenger flow data for scientific processing and analysis, respectively, the holiday passenger flow time distribution, the distribution of passenger payment mode visualisation, mining working days and non-working days change characteristics, and use SARIMAX model for timeseries of metro passenger flow for short-term prediction, and reveal its intrinsic correlation law, followed by random forest regression model for Then, the model is modelled by the Random Forest Regression model, and the final prediction is obtained by using the traversal search combined with the allocation ratio. The performance of the model is evaluated by calculating the average percentage error and prediction accuracy, and the results show that the average percentage error is 25.58%, and the prediction accuracy between the actual value and the predicted value reaches 74.42%, which is significantly better than that of the individual model prediction accuracy.