Determine The Order Of Auto Regression Model Via Fused Lasso

Model for variable selection is an important problem in modern statistics,effectively identify important variables can improve the prediction ability of model.For autore-gressive model of time sequence,when the forecast quantity is big,we in order to filter variables,select the variables.In addition to the variable selection,regression model and order problems,commonly used method with AIC truncation criterion,BIC truncation criterion,but these methods not only to parameter estimation and order can be divided in-to two processes,and retain the smaller coefficient,so we need to improve on the method.Method based on Fused Lasso to set order autoregressive model,the parameter esti-mation and model in order to achieve order at the same time.The purpose of this paper use the LOO cross verify two harmonic parameters,and then on the premise of harmonic parameters given,using Fused Lasso method on the return of the autoregressive model parameter estimation,and by using Monte Carlo method to calculate:???_m1,m2,finally BIC truncation criterion is used to determine the order number of p.using Fused Las-so is much more accurate than Lasso to deal with the self-regression model,which is more time-consuming.Finally,the simulation calculation proves the effectiveness of the method.