Ant colony optimization approach to attribute reduction problem

In order to obtain the minimal reduction of decision table attributes, an attribute reduction algorithm is proposed based on ant colony optimization. The significance of attributes defined from the viewpoint of information theory is used as the heuristic information. The algorithm directly imports the core into each solution constructed by ants and reduces the problem scale. The new state transition rule and pheromone updating rule reflects the orderless characteristic among attributes, and benefits the search in the neighborhood of good solutions. The algorithm is verified on nine typical instances. Experimental results show that, compared with the existing algorithms, the proposed algorithm can find the minimal reduction more easily with less time.