一种基于改进的自组织特征映射网络的文档聚类方法

采用标准的自组织特征映射(Self—OrganizingFeatureMap)网络进行文档分类时存在着收敛速度慢、不同的初始条件及学习样本输入顺序影响学习过程和学习结果等缺点。针对这种情况,论文提出了一种改进的自组织特征映射网络文档分类方法,即采用自适应的方法建立网络拓扑结构,利用输入训练样本来确定网络中的连接权值,综合系统能量函数和训练次数作为判断网络的学习结束标准,并且取得了比较好的实验结果。

A Document Clustering Method Based on Improved SOFM Network

Adopting standard SOFM in document clustering,there are some defects including low speed of constringency,different effects on learning and result from different initialization condition and order of input samples.This paper presents a document clustering method which includes some betterments,such as to set up the network structure with self-adaptive method,to determine the initialization weights on input samples,and to determine the end on both system energy and time of learning.At last,it acquire s better result in experiment.