Improvement of Discrete Particle Swarm classification system

The Discrete Particle Swarm Optimization (DPSO) has little parameters and high convergent capability of the global optimizing. Based on the existing PSO-based classification system we constructed a new classification system based on Discrete PSO. We used the variable-length method to represent particle in the process of operation, represent rule set in a reasonable way and do some appropriate cut, and use the default rule to improve classification effectiveness. The experimental results shown that the system can be cut the rules accurately. It could work well with less number of the rules and desired classification accuracy; the classification system has good performance.