Attribute reduction algorithm for decision table based on relative discernibility matrix

Based on information theory and rough set theory,a new attribute reduction algorithm based on relative discernibility matrix is proposed to solve the problems of the present attribute reduction algorithms.The core attribute is the foundation of this algorithm,through establishing the relative discernibility matrix to reduce the searching room,the condition entropy is used as heuristic information,and gradually add attributes which have the largest condition entropy,until getting the smallest reduction.And the complexity of this algorithm is analyzed.Finally,the experimental results show that this algorithm is effective in attributes reduction of decision tables.