Sparsity Score: A new filter feature selection method based on graph

Mingxia Liu, Dan Sun, Daoqiang Zhang

Abstract: Recently, l 1 graph based analysis using sparse representation has received much attention in pattern recognition and related communities. In this paper, motivated by the success of l 1 graph in dimensionality reduction, we extend it to feature selection and propose a novel filter-type method called Sparsity Score (SS) which ranks features according to their respective sparsity preserving capability. For that aim, a l 1 graph is constructed based on sparse representationon samples, where a l 1 -norm based optimization is used to simultaneously determine the graph adjacency structure and corresponding graph weights of the l 1 graph. Experimental results on a series of benchmark data sets show that the proposed SS method achieves better performance than conventional feature selection methods.