Design of Niche Pso Reduce Algorithm for Hybrid Attributes Based on Neighborhood Rough Sets

A reduce algorithm based on the neighborhood granulation and niche Particle Swarm Optimization (PSO) algorithm is proposed for the reduction of the real decision system with numerical attributes. In this scheme, a rough model is used based on the neighborhood equivalence. The indiscernibility relation is measured by the neighborhood relation, and the universe spaces are approximated by the neighborhood information granules. The use of the niche technology can avoid the preconvergence of the PSO. The select of fitness function and the adaptive across probability are designed, and the reduction algorithm is presented as well. Furthermore, the dependency function is used to evaluate the significance of the numerical attributes. Experimental results demonstrate the validity and feasibility of the proposed algorithm, in application to a classical data set and four University of California at Irvine (UCI) machine learning databases.