Performance Analysis of Cooperative PSO Algorithm with ACO and Tabu Search for Discovering A Rule Set

This work proposes an algorithm for data mining called Cooperative Particle Swarm (PSO)and Ant Colony Optimization(ACO) with Tabu search(TS). The goal of this algorithm is to extract classification rules from data. The algorithm is inspired by both research on the behavior of a group of birds or a school of fish looking for food and real ant colonies. In PSO, when particle finds a local optimal solution, all of the particles gather around the one, and cannot escape from it. On the other hand, TS can escape from the local optimal solution by moving away from the best solution at the present. Real world classification problems are often described by nominal and continuous attributes. ACO based classification algorithm Ant-Miner has the limitation of being able to cope only with nominal attributes in its rule construction process. The PSO based classification algorithms perform well in continuous attributes. The proposed algorithm use cooperative PSO approaches to extract classification rule from mixed data set, use PSO with Tabu search to handle continuous attributes and the pheromone updating of ACO and PSO to handle nominal attributes.. We compare the performance of the cooperative PSO algorithm with the PSO/ACO, data mining algorithm for classification, in eleven public domain data sets. The results provide a evidence that: (a) The proposed algorithm is competitive with existing algorithm with respect to predictive accuracy; and the rule lists discovered by the algorithm are considerably simpler (smaller) than those discovered by the existing algorithm and (b) Reduce the execution time of the algorithm.