Rough Set Attribute Reduction And Its Application In The Analysis Of Investment Environment Indexes

Attribute reduction is one of important aspects of rough set theory. In this paper, we propose a new method for attribute reduction, namely, parallel greedy attribute reduction (PGAR). This algorithm tries to obtain some reductions at one time and these reductions are different each other. After comparing with gene algorithm (GA), particle swarm optimization (PSO) and ant colony optimization (ACO) by using 20 UCI datasets, we can draw the conclusion that the performance of attribute reduction of PGAR is better than that of GA, PSO and ACO. Besides, we also present extended ant colony attribute reduction (EACOAR). The difference between EACOAR and ACOAR is that EACOAR tries to adjust the probability of selecting attributes in the process of reduction construction. By this way, the probability, of which one reduction is the same as other reductions, can be reduced. And the chance of getting the minimal reduction is increasing. The experimental results show that EACO is better than ACO in terms of attribute reduction. At last, we study the importance degree of indexes of inverstment environment by applying rough set positive theory and parallel greedy attribute reduction algorithm. Moreover, according to the analysis of investment environment, we give some suggestions concerning the improvement in investment environment.