摘要
鉴于文本数据具有方向性数据的特征,可利用方向数据的知识完成对文本数据聚类,提出了模糊方向相似性聚类算法FDSC,继而从竞争学习角度,通过引入隶属度约束函数,并根据拉格朗日优化理论推导出鲁棒的模糊方向相似性聚类算法RFDSC.实验结果表明RFDSC算法能够快速有效地对文本数据集进行聚类.
One of the important characteristics of text clustering in datasets is that each cluster center in the dataset has a direction that is different from that of all other cluster centers. This directional information should be incorporated in clustering analysis. In this paper, a new robust fuzzy directional similarity clustering algorithm （RFDSC） is proposed by introducing membership constraints. The new objective function was constructed. Finally, the robustness and convergence of the proposed algorithm were analyzed from the viewpoint of competitive learning. Experimental tests of text clustering in datasets using RFDSC demonstrate its effectiveness.
出处
《智能系统学报》
2008年第1期43-50,共8页
CAAI Transactions on Intelligent Systems
基金
国家"863"资助项目(2006AA10Z313)
国家自然科学基金资助项目(60773206
60704047)
国防应用基础研究基金资助项目(A1420461266)
教育部科学研究重点基金资助项目(105087)
关键词
聚类算法
方向相似性
鲁棒性
竞争学习
clustering algorithm
directional similarity
robustness
competitive learning
