Zip Regression Model Of Variable Selection And Its Application In The Premium Is Determined

Variable selection is an important and popular topic in statistical analysis, and zero-inflation is common in count data. In this dissertation, we provide a variable se-lection approach for Zero-inflated-Poisson(ZIP) regression model. Because of the com-plexity of the ZIP regression model, this problem is not discussed so much. We provide a new variable selection procedure:EM-Adaptive LASSO to solve this problem. The innovations of the achievements in this dissertation are described as following.(1)We provide the EM-Adaptive LASSO variable selection procedure, which is combined with EM algorithm and Adaptive LASSO penalty. And we prove the Oracle properties of the EM-Adaptive LASSO variable selection procedure. The effect of the procedure is compared with other procedures using Monte Carlo simulation.(2)The EM-Adaptive LASSO variable selection procedure is applied to Ratemak-ing. There are16explanatory variables in this model. We select the significant variables for this model using EM-Adaptive LASSO variable selection procedure.