Abstract: A maximum-entropy version of the spherical K-means document clustering algorithm ME-SPKM is presented based on the well-known maximum-entropy. The proposed method uses the cosine similarity which is adopted by the typical text clustering algorithm SPKmeans,then constructs a maximum-entropy-based objective function. Experimental results demonstrate that the maximum-entropy spherical K-means ME-SPKM can achieve better clustering results than traditional clustering approaches in text clustering.

Key words: maximum-entropy, document clustering, spherical K-means

摘要： 提出了一种基于极大熵理论的球面K均值文本聚类算法ME-SPKM。该算法利用了传统文本聚类算法SPKmeans中使用的余弦相似度度量,进而引入极大熵理论构造了适合文本聚类的极大熵目标函数。对文本数据的实验证明了极大熵球面K均值文本聚类算法取得了比传统文本聚类算法更好的聚类效果。

关键词: 极大熵, 文本聚类, 球面K均值