Attribute Reduction Algorithm Research For Rough Set Based On GA-PSO

Solving the minimum attribute relative reduction is an NP problem, there exist serious insufficiency in solving such a problem by using traditional algorithms, therefore, an inevitable trend is to introduce swarm intelligence algorithm. The genetic algorithm will produce different problems in adopting different coding strategy in the process of attribute reduction, thus resulting in unwarranted block assumptions and premature convergence, which will eventually make it difficult to achieve the global optimal. When we use Particle swarm algorithm to process attribute reduction, particle swarm can easily fall into local extreme points during the iteration, leading to the unavailability of global optimal solution and the minimum relative attribute sets. Aiming at these deficiencies, this paper proposes a hybrid intelligent algorithm and uses it to reduce the rough set attribute. The new algorithm can strengthen the local search ability while maintain global optimization, thus can quickly and efficiently get the minimal attribute set.This article firstly introduced the Rough set theory basics including attribute importance, support, etc. based on the research background. More detailed analysis and research for two compared traditional attribution reduction algorithm, The algorithm can reduce the attribution of rough set with the help of distinguishing matrix.Secondly, given the components、operational framework of genetic algorithms. Then the genetic idea be added to the rough set attribution reduction. At the same time, came into another intelligent algorithm-particle swarm optimization algorithm. Join PSO idea into rough set attribution reduction. Also offer an intelligent reduction algorithm to overcome its shortcomings.Next for genetic reduction algorithm premature convergence disadvantage, an improved genetic reduction algorithm focus on optimization strategies of GA. Mainly on aspects of the repair strategies, binary coding initial population, fitness function to optimize the adaptation in genetic algorithm. Put forward to an improved genetic reduction algorithm. PSO can easily reach for local solutions as a weakness of particle swarm optimization. An improved reduction algorithm appeared on its start fitness function optimization.Finally, on the basis of improved particle swarm reduction algorithm, combining the genetic thought. The new algorithm is verified by simulation CTR data and UCI data set. Compare the fitness curve and average uptime on ZOO data sets with other algorithms. Improved GA-PSO attribute reduction algorithm can spend less time on gaining relative attribute reduction than GA attribute reduction algorithm. It can gain the minimum relative attribute reduction at the cost of runtime compared with PSO attribute reduction algorithm. Further prove the effectiveness of the proposed algorithm of this article.