Abstract
AbstractModel selection has become the widely used method to include the relevant predictor variables on the response variable and to remove the irrelevant variables. Bayesian penalized methods such as the elastic net methods are used to address the problem of grouping predictor variables effects. In this paper we concentrated on the sparsity procedure in linear regression model by using elastic net regularization method. We developed Bayesian elastic net by employing the scale mixture of normal distribution mixing with Rayleigh distribution as Laplace prior distribution for the regression parameters. The new scale mixture generates normal mixing with truncated gamma distribution, also we proposed new Gibbs sample algorithm. The proposed Bayesian elastic net method examined by applying real data some and the results shows that the proposed model is comparable with the other regularization methods.
