Research On Attribute Reduction Of Rough Set With Particle Swarm Optimization

Rough Set Theory is an effective mathematical tool to deal with uncertainty,inaccuracy and incomplete information.It has become a research hotspot in intelligent information processing,and has been widely used in many fields such as machine learning,pattern recognition,fuzzy control,fault diagnosis,data mining and knowledge discovery,etc.Attribute reduction is an important content in rough set and it is also an important step for data mining.Traditional attribute reduction can not deal with high-dimensional and complex large data.Therefore,attribute reduction based on particle swarm optimization are studied systematically in this thesis,and typical UCI data and actual oil logging are used to simulate and analyze.The main research contents and innovations are as follows:(1)The traditional attribute reduction is analyzed.Firstly,the basic principles of rough sets,attribute reduction based on the discernibility matrix and attribute reduction method based on attribute importance are illustrated.The results show that the common attribute reduction method can do well in low dimensional small sample information,but the effect is opposite in dealing with high-dimensional and complex large data.(2)The improvement of particle swarm optimization(PSO)algorithm.In order to solve the problem of limited search space,easy to fall into local extreme value and less population diversity,CCQPSO algorithm which is based on quantum,cloud model and catfish effect is studied in this thesis.And cloud model is used to adjust contraction expansion factor to balance global search and local search ability,and then catfish effect is used to change population diversity.The classical test functions show that the CCQPSO algorithm is superior to the PSO algorithm and the quantum PSO algorithm.(3)The attribute reduction based on particle swarm optimization is studied.In order to solve the problem that traditional attribute reduction methods can not deal with high-dimensional and complex data efficiently,the attribute reduction method based on CCQPSO algorithm is studied.Firstly,the mathematical model of attribute reduction are analyzed.And then,the fitness function be constructed which satisfies the reduction condition,the reduced attribute number and its importance degree.The UCI data are tested to show that the attribute reduction algorithm based on CCQPSO algorithm is superior to the traditional attribute reduction algorithm in dealing with high dimension and complex large data.(4)The application of logging oil layer recognition.In order to overcome the drawbacks of low recognition rate of the oil layer in the traditional logging engineering,theattribute reduction based on CCQPSO algorithm and the AdaBoost-CCQPSO-SVM classifier are studied,and the intelligent oil layer recognition system is designed.The actual oil logging data are tested to show that the intelligent oil layer identification system is consistent with the actual test results,and improve accuracy of oil reservoir identification effectively.