Classification Rule Discovering Based On Ant Colony Algorithm

Social insects have high swarm intelligence. Inspired by this mechanism, Ant colony algorithm (ACA) has emerged recently as an evolutionary algorithm or a new metaheuristic algorithm and has been widely applied to solving NP-hard problems combinatorial optimization. Combined with ACO, The paper mainly makes research application on classification problem in data mining (DM). It not only introduces the advantage ACO but also analyze kinds of current data mining algorithms at home and abroad. Accordingly, it presents an algorithm for solving classification problem in data mining.As for the problem of classification rule discovery, the paper also introduces some current classification methods, such as ID3 and C4.5 which is based on the decision tree, BP which is base on Artificial Neural Networks, based on Bayesian statistical classification, classification models and algorithms which are based on ant colony algorithm, such as Ant-Miner Classifier and so on. All these algorithms make solutions for classification problems through classification rules discovery. As for the characteristics of classification problem and the parallel predominance and positive feedback features of ACO, it presents an algorithm for classification rule discovery (ACR). In this algorithm, it uses reasonable probability formulas for ant's attributes and attribute values and employs the strategy for rule quality measurement, therefore, ACR could well discovery the classification rules. Experimental results on several benchmark datasets show that ACR can discover classification rules with better quality.The paper applies ACO to the problems in data mining, especially on classification rule discovery which have wide research prospect. Finally, it makes research on further work, such as ACO on biology databases and other aspects. It also provides a new platform for solving data mining tasks.