Mining Classification Rule Based On Ant Colony Algorithm

An Ant Colony Optimization algorithm is essential a system based on agents that simulate the natural behavior of ants, including cooperation and adaptation. It is collateral,easy realization and combine with other method that An Ant Colony Optimization algorithm is applied more and more widely. Due to it wide application, Classification is an important subject of data mining.This paper proposes an algorithm, which is based on ant colony algorithm, for mining classification rule from categorical database. The main idea is searching a rule by ants, removing samples covered by the rule and repeating this process until get a group of rules by using the strategy of relative probability. Experiment on five public data set shows that compared with CN2 and Ant-Miner, the algorithm can discovered better classification rule, including rule set with better predictive accuracy rate and fewer rules, simpler rule with fewer terms. It proves the algorithm is a better algorithm.