Granular Reduction Based on Boolean Matrix in Formal Decision Contexts

Abstract

Granular reduction is an effective dimension reduction method in formal decision contexts. However, it is a NP-hard problem to calculate the reducts through the discernibility matrix when the data is high-dimensional, which will lead to high time consumption. In this paper, we propose a granular based-matrix reduction in consistent formal decision contexts. First, the consistency of formal decision context is defined based on the object granular matrix. Next, to obtain a minimal granular reduct, we construct the dis-cernibility matrix and reduct algorithm (DARMT). Second, the implication rule approach is studied from the viewpoint of granular based-matrix reduct. Finally, we perform an experimental evaluation on eighteen data sets which explicate the feasibility and effectiveness of our proposed approach.