一种基于SOM的中文文本两阶段快速聚类方法应用研究

研究了一种基于自组织神经网络的中文文档聚类方法,按照提出的中文聚类模型,该方法首先对文档进行SOM训练产生初步的聚类结果,然后对第一次聚类结果进行基于参考点和密度的快速聚类,理论分析和实验结果也证明了该方法适合高维中文文档的聚类,与直接聚类方法相比,该方法提高了聚类的效果,减少了计算时间。

A Two-Stage Fast Clustering Method of Chinese Documents Based on SOM

This paper studies a clustering method of Chinese documents based on self-organizing map neural network(in short SOM).A new Chinese documents clustering model is given.The method first uses SOM to produce a clustering result by learning documents.Then,in the second stage,the clustering results are clustered on reference and density.Both theoretic analysis and experimental results confirm that the method can be used in clustering very large Chinese documents.To compare with other clustering method for Chinese documents,this method performs well and reduces runtime.