An improved artificial bee colony algorithm for the minimal attribute reduction

In this paper, we study the issue of attribute reduction, which is important in rough set theory. However, it has been proved to find the minimal attribute reduction is a NP-hard problem. Numerous heuristic based algorithms have been presented to try to solve this problem. In our paper, we use an improved artificial bee colony algorithm(IABCMR) to find the minimal reduction. We firstly define the significance of an attribute. And then, a selection operator is introduced into the artificial bee colony algorithm to select attributes using the significance as the heuristic information. Thirdly, a crossover is designed as the search strategy of the IABCMR to find more reductions. Finally, several experiments are used to compare the work with some other algorithms, and the results show that this algorithm is effective on big data sets.