A Multiple Evolutionary Neural Network Classifier Based on Niche Genetic Algorithm
Jiang Wu, Changjie Tang, Taiyong Li, Yue Nan Jiang, Shangyu Ye, Xun Lin, Jie Zuo · 2008
Classification is important in data mining. In this paper, a multiple evolutionary neural network classifier based on niche genetic algorithm (MNC-NG) is presented, which establishes classifiers by a group of three-layer feed-forward neural networks with high accuracy and good diversity. The neural networks are trained by niche genetic algorithm based on clustering. The class label of the identifying data can first be evaluated by each neural network, and the final classification result is obtained according to the dynamic voting rule. Experimental results on 6 data sets show that MNC-NG increases the predictive accuracy by 5.6%, 5.5% and 8.5% respectively compared with BP, GA and LM training methods and by 6.0%, 6.1% and 4.0% compared with Naive Bayesian classifier, C4.5 and SVM.