An Efficient Algorithm for Attribute Reduction Based on Discernibility Matrix

A definition for reducing discernibility matrix is provided, and a reduction relative of discernibility set is analyzed to reduce the time complexity of attribute reduction algorithm based on traditional matrix. Use iterative idea to select the necessary condition attributes adding to this reduction relative. Optimize the reduction relative by complete optimization theory and resulting in a Pawlak reduction. Simulation results show the efficiency of the new algorithm.