摘要： 提出了一种用径向基函数(Radial Basis Function)神经网络进行属性约简的方法，定义了影响度等概念来量化评价各属性与待求解问题关系的紧密程度，根据影响度对属性做出约简处理。实验表明，该方法是合理、有效的，明显地提高了机器学习的性能。

关键词: 数据预处理, 属性约简, 径向基函数(RBF)神经网络

Abstract: This paper presents a new attribute reduction method. By using RBF neural network and the defined concepts: influence, people can evaluate different degrees of influence casted by corresponding attribute on results. The experiments on standard data set show that, the new technique using RBF neural network enhances the performance of learning evidently.

中图分类号: 

• TP183