论文部分内容阅读
将奇异值分解同自然正交分解相结合,提出一种改进的正交奇异值分解方法。通过对原始数据进行自然正交分解,削弱原始数据之间的相关性,增强其用于分析及预测的能力,并得到相互正交的主成分代替原始数据进行奇异值分解,分析两个变量场之间的相关关系。在此基础上建立神经网络预测模型,实现多元时间序列的预测。采用该方法对三门峡处径流量同太平洋海温的耦合关系进行分析,并同常规奇异值分解方法进行比较,仿真结果验证了所提方法的有效性。
Combining singular value decomposition and natural orthogonal decomposition, an improved orthogonal singular value decomposition method is proposed. By orthonormal decomposition of the original data, the correlation between the original data is weakened and its ability to analyze and predict is enhanced. The principal components that are orthogonal to each other are replaced by the original data for singular value decomposition. Two variables fields The correlation between. Based on this, a neural network prediction model is established to realize the prediction of multivariate time series. This method is used to analyze the coupling relationship between the runoff at Sanmenxia and the sea surface temperature in the Pacific Ocean and to compare with the conventional singular value decomposition method. The simulation results verify the effectiveness of the proposed method.