基于神经网络和遗传算法的事例最近邻法检索模型

Case K-NN Retrieval Model Based on Neural Networks and Genetic Algorithm

提出了基于自适应共振神经网络ART1模型进行事例的智能层次聚类和基于遗传算法 (GA)进行事例特征权值优化的解决方案。经过ART1网络的层次聚类形成事例库的层次智能存储组织 ,可有效缩小事例的搜索空间 ,提高检索效率。基于GA对特征权值优化可提高检索质量 ,防止检索出的相似度系数最大的事例并非最佳事例 ,即K NN收敛不到最佳解。因此所提方法的运用可有效提高CBR系统整体的检索效率与质量 。

A methodology, in which cases are clustered hierarchically based on ART1 neural networks and feature weights are optimized based on genetic algorithm (GA), is proposed By clustering, the case searching space could be decreased efficiently and the K NN based retrieval quality could be improved greatly by optimizing the feature weights So the methodology presented in this paper could improve the overall retrieval quality and efficiency in CBR systems, especially for those which possess large and complex case library