基于FVSM和自组织映射网络的Web文本自动分类方法

针对Web信息挖掘中的文本自动分类问题，提出了一种基于模糊特征向量(FVSM)和自组织特征映射网络的分类方法．网络由输入层和竞争层组成．输入层节点与竞争层节点实行全互连接．输入层完成分类样本的输入，竞争层提取输入样本所隐含的模式特征，并对其进行自组织，在竞争层将分类结果表现出来．分无监督和有监督两个阶段完成对网络的分类训练．该方法在特征提取时充分考虑了特征项在文档中的Web位置信息，构造出模糊特征向量，使自动分类原则更接近手工分类方法．以中国期刊网全文数据库部分文档数据为例验证了该方法的有效性．

Document automatic classification method base on fuzzy eigenvector and self-organization characters mapping network

Aimed at problems of documents classification in data mining, a classification method is presented based on fuzzy eigenvector and self-organization characters mapping network. The network is constituted of input layer and competition layer whose nodes link with each other totally. The input layer performs classification samples provision, competition layer extracts the implicit pattern characters of input samples and takes self-organization to them, then represents the classification result at competition layer. The network training includes two phases of non-supervisal and supervisal. The feature Web information of its locality in the document is considered while the features are extracted and the fuzzy Eigenvector is constructed, as a result, the automatic classification principle is close to the manual classification method. Finally the availability of the model and algorithms is proved by part documents of china periodical document database.