Multidimensional Correlation Time Series Modeling Method Based on Screening Partial Least Squares Regression of Correlation Variables

Summary: In view of the problems of high dimensionality and serious multiple correlations between variables in time series in many fields, a correlation variable is proposed.Screening Partial Least Squares Regression(CVS-PLSR)Modeling Algorithms.The algorithm introduces correlation-based feature selection(PLSR)As the core algorithm of modeling, it effectively solves the problems caused by multiple correlations between variables.Hazard.The proposed algorithm was verified by using the predicted data of pulp element grade, and the improvedCVS-PLSRThe model obtained by the algorithm is the most compact, and the root mean square error of the test set is only 1.690 2, the correlation between the predicted value and the measured value reaches 97%