基于ARIMA-SVM组合模型的交通流量短时预测

Short-Time Traffic Forecasting Based on ARIMA-SVM Combined Model

随着经济和社会的快速发展,城市交通的拥堵问题日益严重,有效的短时交通量预测可减少盲目出警与控制警力成本。为此,本文采用ARIMA时间序列模型提取道路交通流量序列中线性特征,得到交通流序列预测残差。SVM模型能提取道路交通流量的残差序列非线性特征,对交通流预测残差进行修正。这两个模型组合能有效地提取时间序列全部特征,预测性能更佳。同时采用ARIMA-SVM组合模型对交通流量时间序列数据进行短时预测,其平均绝对误差和平均百分比误差分别为7.7%和12.7%,说明ARIMA-SVM组合模型预测效果较好。在ARIMA、SVM、ARIMA-SVM模型中,SVM模型预测性能最差,其平均绝对误差和平均百分比误差分别为28.2%和46.2%。

With the rapid development of economy and society, the congestion problem of urban traffic is be-coming more and more serious, and effective short-time traffic prediction can reduce the cost of blind police dispatch and control police force. In this paper, the ARIMA time series model is used to extract linear features from the road traffic flow series to obtain the traffic flow series prediction re-siduals, and the SVM model can extract the non-linear features of the road traffic flow residual se-ries to correct the traffic flow prediction residuals. The combination of these two models can effec-tively extract all features of the time series, and the prediction performance is better. The combined ARIMA-SVM model is also used to predict the traffic flow time series data in short time, and its av-erage absolute error and average percentage error are 7.7% and 12.7%, respectively, which indi-cates that the combined ARIMA-SVM model has better prediction performance. Among the ARIMA, SVM, and ARIMA-SVM models, the SVM model has the worst prediction performance, with its aver-age absolute error and average percentage error of 28.2% and 46.2%, respectively.