Binary Ant Colony Algorithm with Particle Optimization Feature for Attribute Reduction Problem

摘要

According to the lack of initial pheromone which leads to the long searching time of the Binary Ant Colony algorithm (BACO), a novel method named Binary Ant Colony Algorithm with Particle Optimization Feature (PBACO) is proposed in this paper. Starting from one-dimensional cellular automata model, the initial solution is produced by Binary Particle Swarm Optimization (BPSO). Then mapping mechanism is introduced to convert the position of particle into pheromone, in this way, a path with distinct discrepancy level of pheromone is generated in the shortest time, and the total searching time of BACO is greatly reduced. At last, PBACO combined with fractal dimension is utilized to solve attribute reduction problem. Six datasets are used to conduct experiments. Experimental results show that our proposed method has comparatively high classification accuracy.