关键词: 城市交通, 短期客流预测, ARIMA算法, 组合改进模型, 常态与非常态客流

Abstract:

Urban rail transit passenger flow characteristics show not only the periodicity, seasonal and normality peak, but also the difference and particularity because of holidays, sports events, urban large-scale events, emergencies, special weather and other factors. In this paper, we carry out the method and realization of more mature normal and poorly studied abnormal passenger flow prediction. Firstly, the general ARIMA time series prediction algorithm is used to analyze the sample history data to realize the normal daily passenger flow forecast. Then for the special factors of passenger flow, we not only put forward a combination model of time series and regression analysis, but also introduce dummy variables and similar daily sample data for further improvement. The scheme realizes the high precision solution of the abnormal prediction problem. The simulation results show that this method has good applicability for the short-term forecasting of passenger flow. In particular, the comparison between the abnormal- state improved combination and the single time series model with the same samples and forecast cycle shows that the improved model can be applied to passenger flow predictions including the inherent nature of time invariant and special factors which has strong adaptability and better prediction accuracy.

Key words: urban traffic, short- term passenger flow forecasting, ARIMA algorithm, combined improvement model, normal and abnormal passenger flow