In this paper, a novel filter-based greedy modular subspace (GMS) technique is proposed to improve the accuracy of high-dimensional data classification. The proposed approach initially divides the whole set of high-dimensional features into several arbitr
