基于SARIMA-SVM组合模型的地铁客流量预测研究

Research on Metro Passenger Flow Forecast Based on SARIMA-SVM Combined Model

为提高地铁客流预测精度,通过结合不同类型日期的地铁客流量周期性及非线性特点,提出融合季节差分自回归移动平均模型(SARIMA)和支持向量机(SVM)的优化模型。该模型采用SARIMA对地铁客流量的时间序列数据进行线性建模,使用SVM对SARIMA模型输入的残差值进行非线性建模,并分别将SARIMA模型、SVM模型及SARIMA-SVM模型对工作日和双休日地铁客流量的预测效果进行比较。实验结果表明,SARIMA-SVM模型的预测精度高于单一模型,相较于不考虑日期类型的组合模型,精度提高了12.24%。考虑日期类型的SARIMA-SVM组合模型能够捕捉地铁客流规律,满足地铁客流量的预测要求,可为地铁运营提供决策依据。

In order to improve the accuracy of subway passenger flow forecast,an optimization model combining seasonal difference autoregressive moving average model(SARIMA)and support vector machine(SVM)is proposed by combining the periodicity and nonlinearity of subway passenger flow on different types of dates. The model adopts SARIMA to carry out linear modeling on the time series data of subway passenger flow,and uses SVM to carry out nonlinear modeling on the residual value of SARIMA model input. The prediction results of SARIMA model,SVM model and SARIMA-SVM model on subway passenger flow on weekdays and weekends are compared respectively. The experimental results show that the prediction accuracy of SARIMA-SVM model is higher than that of single model,and the accuracy is improved by12.24% compared with the combined model without considering the date type. The SARIMA-SVM combined model considering the date type can capture the subway passenger flow law,meet the prediction requirements of subway passenger flow,and provide decision-making basis for subway operation.