关键词: 自组织特征映射, 遗传算法, 文本聚类

Abstract: As a cluster of high-dimensional visualization and unsupervised learning algorithm, Self-Organizing Map (SOM) provided a favorable means for Chinese Web clustering. However, the SOM algorithm has a natural flaw of being sensitive to the network initial power value so as that the accuracy of the cluster made by the SOM has been influenced. To solve the problem, this paper applied genetic algorithm to optimize SOM. This paper made the following contributions. 1) Propose a text clustering method based on GA-SOM-based Text Clustering Algorithms (GSTCA); 2) Make comparison experiment. The result of the experiment shows that the GSTCA has higher accuracy rate than SOM algorithm in the Web Chinese Document Clustering, and the average value of F measure is improved by 14.5% than traditional method. The experiments also show that GSTCA is not sensitive to initial weights of the network, thereby enhancing the stability of the algorithm.

Key words: Self-Organizing Map （SOM）, Genetic Algorithm （GA）, text cluster