【作者】 黄显峰； 方国华； 邵东国；

【Author】 Huang Xianfeng;Fang Guohua;Shao Dongguo;College of Water Conservancy and Hydropower Engineering,Hohai University;State Key Laboratory of Water Resources &Hydropower Engineering Science,Wuhan University;

【机构】 河海大学水利水电学院； 武汉大学水资源与水电工程科学国家重点实验室；

【摘要】 针对水资源要素的时间序列非线性预测难题,建立了基于相空间重构的时间序列混沌预测模型。通过将原始时间序列非线性映射到一个高维特征空间中进行相空间重构,获得预测模型所需要的输入向量和期望输出向量,选择统计学习理论中的SVM模型进行预测,经实证分析,本文模型比ANN模型和AR模型具有拟合效果好,预测精度高,泛化能力强等优点。更多还原

【Abstract】 Aim at the nonlinear forecasting of water resources time series,a forecasting model based on state space reconstruction is established.The original time series is reconstructed to a high characteristic dimension space through nonlinear mapping so as to gain the input vector and anticipant output vector.The SVM model based on statistical learning theory is chosen for the prediction.By example study,the SVM model has such advantages as better curve fitting,higher forecasting precision and stronger generalization.更多还原