摘要
在数据挖掘和知识发现的研究过程中,往往需要对大量的数据进行概念层次分类处理,现广泛使用的方法是通过经验知识或主观判断进行处理,由于知识背景和经验层次的不同,导致不同的人对同一批数据同一问题提出不同的规则知识,加大了对规则知识进行合理性评估的难度。基于此问题,对传统的基于模糊统计分析模型的概念层次分类方法进行改进,首先,通过专家系统获取各模糊样本集,根据离散化的样本属性值及其相对隶属频率值对,进行RBF神经网络训练;然后,利用训练好的神经网络求取各模糊集的最模糊点;最后根据最模糊点获得各模糊集的区域划分,从而实现概念层次的分类处理,避免了传统方法处理过程的复杂性和主观性。
In the process of data mining and knowledge discovery, It often deals with a lot of data according to the concept hierarchies. Now many methods only process data according to empirical strategies or manual estimation. Due to the difference of knowledge background and empirical strategies, different people gained different knowledge on the same data sets and the same issue, which increases the difficulty of evaluation on Me knowledge. This paper modified the traditional model of fuzzy statistical analysis. First we gained the fuzzy sample set by expert system, and trained the RBF NN with the pairs of values of discreted attribute value. Then, worked out the fuzziest point by the trainned NN. Finally, we realized the concept hierarchy classification according to the fuzziest point and avoided the complexity and subjectivity of the traditional method.
出处
《测绘科学》
CSCD
北大核心
2010年第2期117-119,共3页
Science of Surveying and Mapping
基金
河南省杰出青年科学基金资助项目(0512000300)
关键词
模糊集
隶属频率
最模糊点
概念层次分类
RBF神经网络
fuzzy set
fuzzy membershi Pfrequeney
the fuzziest point
concept hierarchy classification
radial basis function NN
作者简介
谢明霞（1985-），女，湖北武汉人，助理工程师，硕士研究生，主要研究方向：空问数据挖掘，GIS。E—mail：ckl702@163．com
