摘要: 特征加权是特征选择的一般情况,它能更加细致地区分特征对结果影响的程度,往往能够获得比特征选择更好的或者至少相等的性能。该文采用自适应遗传算法来优化Category ART网络的特征权值,提出了一种改进的Category ART网络FWART。在UCI标准数据集上的实验表明,FWART网络获得了比Category ART网络更好的泛化能力。将该网络应用在地震震型预报上,取得了很好的预报效果。
关键词:
Category ART神经网络,
特征加权,
遗传算法,
震型预报
Abstract: Feature weighted is the general case of feature selection, which has better performance than (or at least has the same performance as) feature selection. In this paper, feature weighted is employed to improve the classification accuracy of the category ART networks. A novel network named FWART network is proposed, in which the self-adaptive genetic algorithm is used to optimize the weight vector. Experiments on the UCI datasets show that the FWART has better generalization power than Category ART. It is applied to predict the earthquake type. The result is satisfactory.
Key words:
Category ART neural network,
Feature weighted,
Genetic algorithm,
Earthquake type prediction
中图分类号:
丁智国;刘 悦;吴耿锋. 基于特征加权的Category ART网络及应用[J]. 计算机工程, 2007, 33(08): 201-204.
DING Zhiguo; LIU Yue; WU Gengfeng. Feature Weighted Based Category ART Network and Its Application[J]. Computer Engineering, 2007, 33(08): 201-204.