针对短时交通流变化的复杂性与非线性特点,分析了分类回归树模型的建立,包括模型的生长、分裂与剪枝,研究了模型在高速路交通流短时预测中的应用,并对美国波特兰州高速路网的真实交通流量数据进行分析建模。采用RMSE与MAPE误差分析法,将试验结果与传统的交通流预测方法ARIMA模型与Kalman滤波预测模型进行比较。对比结果表明:分类回归树预测模型的RMSE比ARIMA模型与Kalman滤波预测模型分别降低了42.1%、13.1%。 In view of the complexity and nonlinearity of short-term traffic flow, the establishment of classification and regression tree model is analyzed, including the growth, splitting and pruning of the model. The application of the model in the short-term traffic flow prediction is also studied. Real Traffic Flow Data from the Portland Freeway Network in U.S.A. Using RMSE and MAPE error analysis method, the experimental results are compared with the traditional traffic flow forecasting method ARIMA model and Kalman filter prediction model. The comparison results show that the RMSE of classification and regression tree prediction model is reduced by 42.1% and 13.1% respectively compared with ARIMA model and Kalman filter prediction model.