摘要： 提出一种基于非负矩阵分解(NMF)的双重约束文本聚类算法。在正交三重NMF模型中，加入文本空间的成对约束信息和词空间的类别约束信息，将不同的特征词项进行分类。利用迭代规则对原始的词-文档矩阵进行分解，获得文本聚类结果。与多种传统半监督文本聚类算法的对比结果表明，该算法具有较高的聚类精度，能提供更准确和有效的聚类结果。
关键词:
半监督聚类,
非负矩阵分解,
成对约束,
类别约束
Abstract: Non-negative Matrix Factorization(NMF) with dual constraints method for document clustering is proposed. It is based on NMF model with adding of pair-wise constraints on documents and categorization constraints of the words. Iterative rules obtained from the original word-document matrix are decomposed to get document clustering results. Compared with a variety of popular semi-supervised clustering algorithm, the method for document clustering can effectively improve the accuracy of document clustering, and can provide more accurate and efficient clustering results.
Key words:
semi-supervised clustering,
Non-negative Matrix Factorization(NMF),
pairwise constraint,
category constraint
中图分类号:
马慧芳, 赵卫中, 史忠植. 基于非负矩阵分解的双重约束文本聚类算法[J]. 计算机工程, 2011, 37(24): 161-163.
MA Hui-Fang, DIAO Wei-Zhong, SHI Zhong-Zhi. Dual-constraints Text Clustering Algorithm Based on Non-negative Matrix Factorization[J]. Computer Engineering, 2011, 37(24): 161-163.