Study on discretization in rough set via modified quantum genetic algorithm

Abstract

Optimal discretization of real valued attributes in rough set is a problem of NP-complete. To resolve this problem, a modified quantum genetic algorithm (MQGA) and a new parametric configuration scheme for the fitness function are proposed in this paper. In MQGA, a novel technique with locally hierarchical search ability is introduced to speed up the convergence of QGA. With this configuration scheme, it is convenient to distinguish the appropriate solutions that partition the new decision table consistently from all the results. Experiments on dataset of Iris have demonstrated that the proposed MQGA is more preferable compared with the traditional GA-based method and QGA based method in terms of execution time and ability to obtain the optimal solution. © 2008 Springer-Verlag Berlin Heidelberg.