Ant Colony Algorithm Is Applied Research In Data Mining

Ant Colony Algorithm(ACO) is a Swarm Intelligence algorithm which shows many promising characters and has been applied successfully to categorical problem. In this paper, we present an improvement to Ant Miner(we called AMI). The algorithm consists of five steps, comprising rule construction, rule pruning, pheromone updating, update convergence test, and removing cases. With the combination of Genetic Algorithm(GA) and ant algorithm, we present a mutation when constructing a classification rule. It expends the range of search by ants effectively; Based on the study of pheromone updating, we applied an evaporation to the algorithm, and improved the method of pheromone updating. And the length of the rule been searched is a parameter in the formula of pheromone updating also. More amount of pheromone will be increasing with each term occurring in a shorter rule by an ant than those occurring in a longer rule.The algorithm was tested and performed better than the original Ant Miner algorithm. It has higher predictive accuracy. And the rule lists discovered by the new algorithm are simpler than those discovered by the original algorithm.