AN ADAPTIVE SEARCHING OPTIMAL ALGORITHM FOR THE ATTRIBUTE REDUCTS

Attribute reduction is one of the key problems for the knowledge acquisition. Based on the rough set theory, the relative difference comparison table is constructed to effectively and efficiently achieve the better attribute reducts. Then the relative difference comparison table is combined with the heuristic knowledge to design three algorithms respectively: the improved algorithm for attribute reducts （AR1）, the complete algorithm for judgement of attribute reduct （RJ）, and the enhanced algorithm for attribute reducts （AR2）. The adaptive searching optimal algorithm for the attribute reducts （ADSOA） is designed by combining the main thoughts of the gene algorithm and the concrete operations of the stimulated annealing strategy with the above three algorithms. And the ADSOA is used to reduce the rheumatoid arthritis diagnosis decision table in the traditional Chinese medicine. The experimentation results show that the ADSOA can obtain the better attribute reducts more effectively and efficiently, and even can achieve the optimal attribute reducts for some specific problems. At the same time, it is concluded that the presence of relative difference comparison tables is practically meaningful for further constructing much more effective and efficient algorithms for the attribute reducts.