Information System Continuous Attribute Discretization Based on Binary Particle Swarm Optimization

Information system discretization widens the use of rough set theory. Rough set attribute discretization should maintain the consistency of knowledge base classification with less cut-points. This paper introduces within the context of rough set a novel method of attribute discretization by applying binary particle swarm optimization (PSO). A minimal set of cut-points are determined with PSO, which is especially adept at global search, while the consistency of knowledge base classification is assured from the high quality of clustering results from rough set theory. The effectiveness of this discretization method is verified in a simulated example of evaluating the vibroacoustic diagnosis symptoms.