Research And Application Of Random LASSO Post-selective Inference Algorithm

The selection method of penalty coefficient variables such as Lasso does not consider the result of the selected event as a statistical behavior when selecting the model,which means the corresponding statistical inference can not be made.Existing literature methods discard information about variables that are not included in the model when selectively inferring the model.In order to solve the above problems,this paper proposes an estimation theory based on adaptive Lasso selective inference,and builds a new algorithm framework based on this.We first use the randomized Lasso algorithm to obtain the importance metric of each variable.That is,the size of the corresponding weight estimate;next,starting from the adaptive Lasso selection event,the corresponding truncated normal parameter distribution is constructed,and hypothesis testing is performed on the distribution to further delete the insignificant variables in the coefficient,so as to obtain the final estimation result.The innovative point of the algorithm proposed in this paper is that,starting from the perspective of the posterior selective inference event,combined with the randomized Lasso algorithm,a trade-off solution is obtained from a new perspective of comprehensively considering data information and selective inference information.The method breaks through the limitation of the number of adaptive Lasso selection variables.When the redundant information of the disturbance variable is introduced into the observation data,it can flexibly carry out feature screening and identify the real variable and the disturbance variable.In this paper,a simulation study of the proposed algorithm is carried out in a variety of situations.Compared with the existing penalty variable selection methods,the numerical simulation results show that our method has a good performance in variable selection and accuracy rate identification.In the analysis of actual economic data,this article uses the proposed algorithm to discuss the actual factors that affect the exchange rate,and compares it with the traditional econometric algorithm.The conclusion is that the algorithm in this paper can correctly select the real variables.It explains the fluctuation of exchange rate well,which is consistent with related economic theory.