基于相空间重构与LSSVM的交通流量预测

针对仅利用欧氏距离不能准确反映相空间中相点间的相似性大小,提出一种改进预测模型,该模型同时考虑相点间的欧氏距离和相似性来选取邻近点。在对交通流量时间序列进行相空间重构后,运用最小二乘支持向量机分别对不同方法得到的邻近点进行训练,并对未来时段的交通流量进行了多步预测。实际案例的预测结果表明,改进方法比一般方法具有更好的适应能力和预测精度。

Traffic Flow Prediction Based on Phase Space Reconstruction and Least Squares Support Vector Machines

An improved prediction model was proposed due to that using only the Euclidean distance cannot accurately reflect the similarity between phase points,when selecting neighboring points after the phase space reconstruction of traffic flow time series.The model took into account the Euclidean distance and the similarity between phase points to select the neighboring points.These selected neighboring points were trained by LSSVM,and a multi-step prediction for the traffic flow of next times was carried out.Prediction results of an actual case show that the improved method is better than the general method in adaptability and prediction accuracy.