Energy Production Analysis Based On Compound Regularization Method

In the statistical regression analysis,it is very important to select the appropriate design matrix.On the one hand,it is necessary to select the factors as small as possible to achieve the simplified model to reduce the amount of the calculation.On the other hand,the factors within the design matrix should be orthogonal to avoid coincide.The choice of factors is to explain the dependent variable from different perspectives,but the explanatory factors are independent.A large number of scholars have done a lot of pioneering work on the choice of variables and collinearity problems.In particular,Hoerl has proposed the ridge regression which can reduce the collinearity of the design matrix.Hoerl considers that the partial regression coefficients are inflated,then the collinearity is reduced by adding a penalty term for the coefficients.In 1996,Tibshirani proposed the lasso method,which adds a penalty term similar to the ridge regression method.The lasso method can play the role of variable selection.The Elastic Net method is an effective improvement of lasso,which not only combines the advantages of ridge regression and lasso method,but also reduces the design matrix collinearity while achieving the purpose of selecting variables.The compound regularization method(Elastic Net)is applied to the prediction of energy field,and some superior properties of the compound regularization method are illustrated by comparing with the ordinary least squares regression(OLS)method.The contents of this article are as follows:The first part briefly reviews the develop-ment of the problem and gives some basic knowledge.The second part,Some properties and calculation methods of general collinearity method are described.In the third part,A brief introduction to Elastic Net method definition and parameter estimation process.In the fourth part,In view of the energy problem,the gasoline production in the mainland of China is chosen for empirical analysis.The fifth part is the summary and discussion of this thesis.