Neural Network Ensemble Based Ant Colony Classification Rule Mining

Generalization ability and comprehensibility are very important for classification rule mining. This paper proposes an improved ant colony classification rule mining method named ant-classifier with strong comprehensibility. Neural network ensemble is with good generalization ability. A novel algorithm NeAnt is also proposed which integrates neural network ensemble and ant-classifier´s merit. In the NeAnt, neural network ensemble is used for preprocessing the original training set and ant-classifier is used for mining rules from the new training set. Experiments show that the generalization ability of NeAnt and ant-classifier can be better than NEC4.5 and C4.5 respectively