演化追踪法优化相空间的SVM供水量预测模型

Improvement of SVM regression forecast water supply model based on phase space reconstruction

相空间重构的支持向量机预测模型应用十分广泛,在城市供水量预测方面也占据着重要地位,传统的预测模型趋向于将重构的相空间整体带入,这样可能存在引入无效相点从而影响预测精度的问题,基于此将演化追踪法引入相空间重构的预测模型对有效相点进行筛选,优化预测模型的训练样本,达到提高预测精度目的。利用MATLAB编程软件将演化追踪法用于城市供水量的预测,预测结果的平均绝对误差由0.52%降低到了0.29%,证明了演化追踪法的可利用性与有效性。

Nowadays,SVM prediction model based on phase space reconstruction is widely used,and it also plays an important role in urban water supply prediction.The traditional prediction model tends to bring the reconstructed phase space into the whole,which may lead to ineffective introduction of SVM.Phase prediction method is used to improve the accuracy of prediction.Based on this,the evolutionary tracing method is introduced into the prediction model of phase space reconstruction to filter the effective points and to optimize the training samples of the prediction model.The evolutionary tracing method is used to forecast the urban water supply quantity by using MATLAB programming software.The average absolute error of forecasting result is reduced from 0.52%to 0.29%,which proves the availability and effectiveness of evolutionary tracing method.