作者投稿和查稿 主编审稿 专家审稿 编委审稿 远程编辑

计算机工程 ›› 2010, Vol. 36 ›› Issue (23): 180-182. doi: 10.3969/j.issn.1000-3428.2010.23.059

• 人工智能及识别技术 • 上一篇    下一篇

属性重要性评分方法的改进

杨宝华,辜丽川,李绍稳,金秀   

  1. (安徽农业大学信息与计算机学院, 合肥 230036)
  • 出版日期:2010-12-05 发布日期:2010-12-14
  • 作者简介:杨宝华(1974-),女,副教授,主研方向:粗糙集,数据挖掘,人工智能;辜丽川,讲师、博士研究生;李绍稳,教授、博士生导师;金秀,助教
  • 基金资助:
    国家自然科学基金资助项目(30800663);国家科技支撑计划基金资助项目(2009BADC4B02);安徽省高校省级自然科学研究基金资助项目(KJ2007B158,KJ2008B111)

Improvement on Attribute Significance Evaluation Method

YANG Baohua,GU Lichuan,LI Shaowen,JIN Xiu   

  1. (School of Information and Computer, Anhui Agriculture University, Hefei 230036, China)
  • Online:2010-12-05 Published:2010-12-14

摘要: 研究信息系统的属性重要性评分方法,通过引入敏感系数构建神经网络模型,提出属性重要性评分算法,将信息系统的各条件属性和决策属性构造一个径向基函数(RBF)神经网络。经训练和学习后,综合考虑各属性间的关系,动态调整RBF网络的拓扑结构,评分各属性的重要性。以红籽西瓜性状数据作为样本数据和测试数据进行实例分析,验证该方法的有效性。

关键词: 敏感系数, 属性, 神经网络, 径向基函数

Abstract: Method for attribute significance evaluation is researched in information system. Attribute significance evaluation algorithm is proposed based on sensitivity coefficient and RBF neural network. Information system about condition attributes and decisionmaking attributes is constructed as a RBF neural network. Attribute significance is determined by RBF neural network training and learning, which analyzes sensitivity between network output and the input. After dynamic adjustment of RBF network topology and judge importance of attribute, the method based on sensitivity coefficient and RBF neural network is proved through the example of redseed watermelon, which shows the algorithm is effective.

Key words: sensitivity coefficient, attribute, neural network, radial basis function

中图分类号: