论文部分内容阅读
PCCS是为了帮助 Web用户从搜索引擎所返回的大量文档片断中筛选出自己所需要的文档 ,而使用的一种对 Web文档进行快速聚类的部分聚类方法 :首先对一部分文档进行聚类 ,然后根据聚类结果形成分类模型对其余的文档进行分类 .采用交互式的一次改进一个聚类摘选的聚类方法快速地创建一个聚类摘选集 ,将其余的文档使用 Nal¨ve- Bayes分类器进行划分 .为了提高聚类与分类的效率 ,提出了一种混合特征选取方法以减少文档表示的维数 :重新计算文档中各特征的熵 ,从中选取具有最大熵值的前若干个特征 ;或者基于持久分类模型中的特征集来进行特征选取 .实验证明 ,部分聚类方法能够快速、准确地根据文档主题内容组织 Web文档 ,使用户在更高的主题层次上来查看搜索引擎返回的结果 ,从以主题相似的文档所形成的集簇中选择相关文档
PCCS is used to help Web users from the search engine returns a large number of pieces of the document screening out the documents they need, and use a fast clustering of Web documents part of the clustering method: first part of the document clustering, And then classify the remaining documents according to the clustering results.Using an interactive clustering method that improves a clustering excerpt at a time to quickly create a clustering excerpt set and use the Nal¨ve-Bayes classification for the rest of the documents In order to improve the efficiency of clustering and classification, a hybrid feature selection method is proposed to reduce the number of dimensions of document representation: recalculate the entropy of each feature in the document and select the first several features with the maximum entropy value; Or based on the feature set in the persistent classification model.Experiments show that some clustering methods can quickly and accurately organize Web documents according to the subject content of the documents and enable users to view the results returned by search engines at a higher topic level, Select the relevant documents from the cluster formed by the similar theme documents