High dimensional feature selection under interactive models

In contemporary statistics, the need to extract useful information from large data boosts the popularity of high dimensional feature selection. High dimensional feature selection aims to select relevant features from the suspected feature space by removing redundant features. Among high feature selection studies, a large number have considered main effects only, although interactive effects are also necessary for the explanation of the response variable. In this study, we firstly expand current studies of the new model selection criterion EBIC (Chen and Chen, 2008) to interactive models and .explore its selection consistency. With the application of EBIC, we develop a novel feature selection procedure, called sequential L1 regularization algorithm (SLR), under high dimensional space by considering both the main effect features and the interactive effect features in the context of generalized linear models. The theoretical property of SLR is explored and the corresponding conditions required for its selection consistency are identified.

Author dc:creator

• HE YAWEI

• feature selection, EBIC, penalized methods