Research On Attribute Reduction Based On Particle Swarm Optimization And Variable Precision Rough Sets

Rough set is a new mathematical tool for dealing with fuzzy and uncertain knowledge,through the set of upper and lower approximations,the information in the system is expressed in the form of decision rules.Due to the limitations of the decision-making of the classical rough set model,it is easy to cause the loss of potentially effective information in the processing of data.For its shortcomings,the variable precision rough set expands the standard inclusion relationship in rough sets to the majority inclusion relation by introducing inclusion degree,which enhances the processing ability of noise data.Attribute reduction is one of the most important contents in the theory of rough sets and variable precision rough sets,that is,the process of removing redundant attributes when the original information system's classification ability remains unchanged.Solving the minimum attribute reduction proved to be an NP problem,it is an inevitable trend for the research of attribute reduction algorithm to seek both to improve the efficiency of the algorithm and to obtain the minimum attribute reduction.This paper proposes two attribute-based algorithms for attribute reduction of variable-precision rough sets.Its main work is as follows:1.Attribute reduction based on tolerance matrix and variable precision rough set needs to calculate the attribute dependency of all attribute combinations,the large amount of calculation leads to low efficiency.Based on this that an attribute reduction algorithm based on improved tolerance matrix and variable precision rough set is proposed,at the beginning of the algorithm,the condition attribute set is set as the minimum attribute reduction,without traversing all the condition attribute combinations and solving the attribute dependency degree,only the condition attribute with fewer number of condition attributes than the current minimum attribute reduction is calculated.Compared with the original algorithm,the efficiency of the solution has been improved.2.Discrete Binary Particle Swarm Optimization combined with rough sets for attribute reduction,and its fitness function has certain limitations,it is impossible to find the minimum attribute reduction when the minimum attribute is reduced to the attribute set itself.In order to improve the precision and efficiency of the reduction,an attribute reduction algorithm based on discrete particle swarm and variable precision rough set is proposed.The algorithm improves the original fitness function,and uses the attribute dependence as a judgment basis,so that it can be automatically adjusted along with the evolution of discrete binary particle swarms to ensure convergence speed and evolution direction,so that an optimal solution,the minimum attribute reduction,can be obtained in the end.Minimum attribute reduction.Through experimental verification,the reduction efficiencies of the two attribute reduction algorithms have been significantly improved.