基于簇特征的文本增量聚类研究

提出了一种基于簇特征的文本增量聚类算法:充分利用简单、有效的k-means算法来进行初始聚类,并保留聚类后每个簇的簇中心、均值、方差、文档数、3阶中心矩和4阶中心矩作为该簇的簇特征,当出现新增数据时,利用初始簇的簇特征对新增数据进行聚类.在20newsgroups数据集上的实验结果表明:相比于对整个数据集进行重新聚类,该算法具有一定的优势.

A Research on the Text Incremental Clustering Based on Cluster Features

A text incremental clustering algorithm based on cluster features has been presented.Firstly,initial clustering is performed by making full use of simple and efficient k-means algorithm.Secondly,the clustering center,mean,variance,the number of document,the third central moment and the fourth central moment are saved as the cluster features of each cluster.Finally,when new documents occur,they are incrementally clustered with those cluster features.The experimental results on 20newsgroups data set demonstrate that the algorithm the paper presents has some advantages.