一种成对约束限制半监督文本聚类算法王纵虎1,2，刘速2（1.中国人民.doc

作者简介：王纵虎(1984-)，男，中国人民大学在站博士后，中国石油规划总院工程师 . 研究方向：数据挖掘，大数据分析. E-mail： HYPERLINK mailto:zonghuwang@ zonghuwang@ 一种成对约束限制半监督文本聚类算法 王纵虎1,2，刘速2 （1.中国人民大学 统计学院 北京 100872）1 ；(中国石油规划总院 计算机信息中心 北京102206）2 摘要：半监督聚类能利用少量标记数据来提高聚类算法性能，但大部分文本聚类算法无法直接应用成对约束等先验信息。本文针对文本数据高维稀疏的特点，提出了一种半监督文本聚类算法。通过将成对约束信息经过扩展后嵌入文档相似度矩阵，在此基础上根据已划分与未划分文档之间的统计信息逐步找出剩余未划分文本集合中密集的且与已划分聚类中心集合相似度较小的K个初始聚类中心集合，然后将剩余的相对较难区分的文档结合成对约束限制信息划分到K个初始聚类中心集合，最后通过融合成对约束违反惩罚的收敛准则函数对聚类结果进一步优化。算法在聚类过程中自动确定初始聚类中心集合，避免了K均值算法对初始聚类中心选择的敏感性。在几个中英文数据集上的实验结果表明提出的算法能有效地利用少量的成对约束先验信息提高聚类效果。 关键词：聚类; 半监督; 向量空间模型；成对约束；文本 中图法分类号： TP391 文献标识码　A Pairwise Constrained Semi-supervised Text Clustering Algorithm WANG Zonghu1,2, LIU Su2 (1 School of statistics Renmin University of China, Beijing 100872)2 Computer Information Center,Petrochina Planning and Engineering Institute Beijing 102206) Abstract Semi-supervised clustering can use a small amount of tag data to improve the clustering performance，but most of the text clustering algorithm can not directly apply priori information such as pairwise constraints. As the characteristics of text data were high-dimensional and sparse, we proposed a semi-supervised document clustering algorithm. First, pairwise constraints were expanded and embedded in the document similarity matrix, then gradually search K density regions which have a small similarity with the already partitioned text collection in the remaining unpartitioned text collection as initial centroid. The remaining unpartitioned texts which were relatively difficult to distinguish are assigned to the K initial centroid according to the constraints. Finally the clustering result was optimized by the convergence criterion function by integration of punish violations of pairwise constraints . In the clustering process, it can automatically determines the initial centroids to avoid the sensitivity to the initial centroids of K-means algorithm. Experimental results show that the pro