The Applied Analysis Of Binary Quantile Regression With Adaptive LASSO

Quantile regression gives a full insight of data information,so quantile regression for binary response data can describe and predict the propensity to make a choice more accurately.With the technology of data collection and storage developing,data information is booming.It is a vital precondition and key for modeling to select significant ones from numerous variables.This paper will revolve variable selection and application of binary quantile regression to conduct the deep research.The main work of this paper is as follows:For the first,we systematical introduce the theoretical basis of binary quantile regression,including the model definition,parameters estimation and prediction methods.Some further improvement for the model are also suggested.And discuss the advantages and disadvantages of variable selection methods,choosing an appropriate way to variable selection in binary quantile regression.For the second,binary quantile regression model with the adaptive LASSO penalty is studied.The core is to choose a suitable prior for parameters,construct a Bayesian hierarchical model and deduce the posterior distributions of parameters.Furthermore,An efficient Gibbs sample algorithm to estimate parameters is proposed.Numerical simulations indicate that the proposed method has a good performance in variable selection and classification.In the end,we apply binary quantile regression model with the adaptive LASSO penalty to the personal credit scoring using the German credit dataset.The proposed method can identify the significant variables via the adaptive LASSO penalty and give a much more detailed insight in the heterogeneous effects of the attributes on personal credit,so it can help crews to assess their client's credit risk pointedly.Then,we compare the proposed method with other models.The results demonstrate that binary quantile regression model with the adaptive LASSO penalty has a good performance on the personal credit evaluation.