相空间重构和SVR联合优化的短时交通流预测

短时交通流预测首先重构相空间，然后采用时间序列模型预测交通流量，而支持向量回归机（SVR）是比较好的时间序列预测模型。但短时交通流相空间重构的嵌入维数与延迟时间与支持向量回归机的参数确定往往是分别独立地求解，难以达到两组参数值的同时最优，影响预测的准确性。为了提高短时交通流的预测准确性，提出一种利用粒子群算法联合优化相空间重构和支持向量回归机的预测模型，并用于实际短时交通流数据的预测。该模型的相空间重构和支持向量回归机（SVR）的参数联合一起优化，利用粒子群算法同时优化其两组参数的组合值。采用短时交通流数据仿真，结果表明联合一起优化所得参数的预测器提高了简单模型预测的效果。

Short-term traffic flow prediction model of phase space reconstruction and Support Vector Regression with combination optimization

Predicting short time traffic flow needs phase reconstruction at first, and then the traffic flow is computed with prediction model of time series. Support Vector Regression（SVR）is a popular forecasting model that has better generality with more powerful theoretical base. However, the embedding dimensions and time delay of phase reconstruction and the parameters of SVR are computed independently, which is difficult to get the optimal parameters in the meanwhile so that accuracy of prediction is not good. In order to improve the accuracy of prediction, a method of short time prediction is proposed which uses combination of phase reconstruction and Support Vector Regression. In this model, the parameters of phase reconstruction and Support Vector Regression are optimized with combination using the PSO. The experiment conducted by traffic dataset has shown that the new method improves the performance of short-term traffic flow forecasting model.