DMK算法在中文文本聚类中的应用  

Application of the DMK algorithm to the Chinese text clustering

作　　者：季圣洁 葛万成[1] Ji Shengjie;Ge Wancheng(Sino-German College,Tongji University,Shanghai 200092,Chin)

机构地区：[1]同济大学中德学院,上海200092

出　　处：《信息通信》2018年第7期1-4,共4页Information & Communications

基　　金：上海市科学技术委员会科研项目<基于个性化推荐技术的航空移动社区服务模式研究与应用>(项目号:14DZ1101400)

摘　　要：对中文文本聚类进行了研究,将所改进的DMK算法应用于实际的中文文本聚类中。将百度百科数据集中的不同类型词条内容经过文本处理(分词、去停用词、特征选取、降维)后分别使用原始K-means聚类算法和DMK(Density-based and Max-min-distance K-means)算法进行聚类,并选择F-measure值及RI值(Rand Index)等指标对聚类结果进行分析。结果表明,针对实验中使用的百度百科中文数据集,DMK算法的F-measure值较原始算法平均提高0.342%,RI值较原始算法平均提高9.34%,验证了所设计的DMK算法对实际中文文本聚类的实质性优化。To research on the Chinese text clustering, applying the DMK（Density-based and Max-min-distance K-means） algorithm to the actual Chinese text clustering. After text treatment（words-splitting, stopwords-removing, feature selection, dimensionality reduction）, different types of article contents from Baidu Encyclopedia was clustered using the original K-means algorithm and the DMK algorithm. The F-measure and Rand Index values were selected to analyze the clustering results. The results showed that, for the Chinese data set from Baidu Encyclopedia used in the experiment, the F-measure value of the DMK algorithm improved an average of 0.342% and the RI value of DMK algorithm improved an average of 9.34% than the original algorithm. The results verified the DMK algorithm＇s effect of improving performance in actual Chinese text clustering.

关 键 词：K-MEANS DMK算法 文本聚类 分词 文本挖掘 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]