Research On Attribute Reduction Method Based On Particle Swarm Optimization

Particle swarm optimization(PSO),which is proposed by Kennedy and Eberhart,.As an swarm intelligence algorithm,PSO has advantages such as simple,practical,fast convergence,strong search ability and so on..Rough set theory,based on indiscernibility relation,can effectively solve data which contain different noises.Attribute reduction theory,which is to remove redundant information in the decision-making table and keep attributes in decision information table dependency unchanged,is good for people to get useful information in big and complex data.Nowadays attribute reduction plays an more and more important role in our society,and it has been successfully applied in many aspects,make a profound impact in our daily life.This paper discusses the theory of particle swarm optimization and rough set attribute reduction method,put a new binary PSO algorithm into attribute reduction method.compared to some other former methods,The new method has better efficiency and performance,demonstrates its value for data preprocessing.The main research work includes the following aspects:1)By analyzing the treatment about attribute value in information system and its advantages and disadvantages,a comparison among existing reduction algorithms has been made and we come to a conclusion.2)Introduced standard PSO and RSPSO,proposed a appropriate optimization algorithm which demonstrates a better search ability and convergence capability.3)Make a binary conversion on the basis of the improved PSO algorithm,then proposed binary attractor weight change regional shock search embedded particle swarm optimization algorithm(BWMRSPSO).4)Form a new method of combining BWMRSPSO with rough set theory to solve the problem of attribute reduction in compatible decision information table.In accordance with dependence of the decision attribute on condition attribute,combined with similar approximate classification accuracy and approximate classification quality forms particle fitness function to guide the rough set attribute reduction and finally obtained a minimum attribute reduction.Finally,experiment proves the effectiveness and practicality of the attribute reduction algorithm proposed.