The search of the attribute reduction algorithm of rough set in incomplete decision table is a research hot spot. Though analysis of the advantages and disadvantages of the existing attribute reduction algorithms,we put forward a definition of relative discernibility matrix base on the positive area. Then we compute the tolerance class with the the idea of cardinal number sorting method, giving a quick heuristic algorithm of attribute reduction with theconditional entropy and relative discernibility matrix, which of the time complexity is in the worst case. The test result shows that the algorithm can obtain an attribute reduction efficiently.
Granular-conditional-entropy-based attribute reduction for partially labeled data with proxy labels
