基于组合模型的城市轨道交通短时客流预测

针对城市轨道交通短时客流的非线性分布特征，本文提出一种基于变点模型、小波变换、自回归滑动平均模型(ARMA)的组合预测模型.首先，利用变点模型将车站进站客流数据划分为具有不同特征的时间段；然后，使用自相关和偏自相关分析确定时间序列的平稳性；之后，分别采用 ARMA模型与小波 ARMA组合模型对北京市某地铁站的进站量进行客流预测，并对预测结果的误差进行了比较分析.经过对比分析表明，小波 ARMA组合模型能够较好地预测出未来的短时客流，预测效果优于单一 ARMA模型，计算速度也能够满足短时预测的需求，该方法可为城市轨道交通的运营组织提供参考建议.

Short-term Passenger Flow Prediction for Urban Railway Transit Based on Combined Model

For short-term passenger flow in urban railway transit has the characteristics of nonlinear distribution, a combined forecasting model based on variable- point model, wavelet transform, and autoregressive moving average model is proposed. The model firstly uses variable- point model to divide the state interval, then uses autocorrelation and partial autocorrelation analysis to determine whether the time series is suitable for the ARMA model. Finally, ARMA single model and wavelet ARMA combined model are used to predict the passenger flow of a subway station in Beijing, and the error of the prediction results is compared and analyzed. The comparison shows that the prediction effect of the wavelet ARMA model is better than that of a single ARMA model, and the calculation speed can also meet the needs of short- term prediction. This method can provide reference for the operation organization of urban rail transit.