Multiple Time Series Correlation Extraction and Prediction Based on EOF-SVD Model
Min Han, Decai Li · Jisuanji fangzhen · 2008
An improved SVD method was introduced by combining empirical orthogonal function (EOF) with singular value decomposition (SVD).By using EOF to original data,the correlation of it could be impaired and its ability of analysis and prediction was improved,the principle components which were orthogonal from each other to substitute original data were received and the correlation between two variable fields was analyzed.Based on this correlation,the neural networks model was established to achieve multivariate time series prediction.Applying the method ,the correlation between runoff yield and sea surface temperature was analyzed.Comparing to traditional SVD,simulations prove the ability of the method.