An Attribute Reduction Method Based on Improved Particle Swarm Optimization Algorithm

Abstract

Minimum attribute reduction is a NP-hard combinatorial optimization problem. Aiming at the attribute reduction problem in the context of high-dimensional big data, a Levy flight particle swarm optimization algorithm based on premature judgment mechanism (LPSO) is proposed. For the particle swarm optimization algorithm is prone to premature problems and the shortcomings of common precocious judgment methods, a more perfect precocious judgment and coping mechanism is designed. Once precocious, the Levy flight mechanism is introduced immediately, and the worst position of the individual is introduced in the algorithm design, which makes the precocious particles fly to the original small probability search space, and the search area is more uniform, thus improving the global search performance, and then the proposed LPSO algorithm is applied to the solution of the minimum attribute reduction problem. The UCI data set test shows the attribute reduction method based on LPSO algorithm method is superior to other reduction methods.