Logistic Model Research Based On Mutual Information Measure And Elastic Net

Elastic Net have been proved to be very effective in variable selection.The existing models with Elastic Net penalty ignore the interaction between variables due to too much emphasis on sparsity.Therefore,we have added a new regular term to models with Elastic Net penalty to emphasize higher-order interactions between variables.First of all,we constructed the hypergraph of features.Each node of the hypergraph corresponds to the related variables,and each hyperedge has a weight,which represents the size of the information of the interaction between variables connected by the hyperedge.In this paper,we use MII(multidimensional interaction information)to measure the weight of the hyperedge.Secondly,we use the feature hypergraph as a regularizer on the covariate coefficients which can automatically adjust the relevance measure between a covariate and the response by the interaction weights obtained from hypergraph.Finally,the alternative direction multiplier method(ADMM)was used to solve the Logistic model based on mutual information meature and Elastic Net.The case analysis on different data sets can prove the effectiveness of our new model.