一种改进的文本聚类中心选择算法的研究.pdf

摘要 K-means聚类算法是一种基于划分的聚类算法，在众多聚类算法的性 能比较中，该算法具有效率高、可发现任意聚类形状、对数据输入顺序不 敏感、且对于高维数据也有较好的性能，而得到广泛的应用，但是K-means 算法需要事先指定K值，并且对于“噪音”和孤立点数据敏感。因此针对 K-means算法的以上缺点提出了新算法，从以下两点对传统的K-means算法 进行了改进。首先，将文档集合分成一定规模的子文档集，串行聚类。其 次，在聚类中心的选择上使用遗传算法的选择、变异操作确定。并且通过 实验证明了，该算法无论是在聚类中心选择的精确性还是聚类结果的准确 性都得到了提高。 关键词：文本聚类；聚类中心；K—Means算法；遗传算法 I ABSTRACT K-means clustering algorithm is a division based on the clustering algorithm, many clustering algorithm in the performance comparison, the algorithm is efficient, can be found clustering of arbitrary shape, the order of data input is not sensitive, and the high-dimensional data also have better performance, so it is used widely. But K-means algorithm need to specify K-value in advanced, and it is sensitiv e to the noise and isolating data points. This article developed a new al gorithm over the shortcomings of the K-means algorithm. Firstly, samples co llected and divided into subsets, serial clustering. Secondly, genetic algorithm will join the K-means algorithm, to select the center of the cluster. And experiments show that, this new algorithm is better not only in the precision of cluster result, but also in the accuracy of cluster center choosing than the traditional K-means algorithm. Keywords: Text clustering; Center of text clustering; K-Means Algorithm; Genetic Algorithm II 目 录 1 绪论1 1.1 引言 1 1.2 国内外研究现状 2 1.3 课题研究的目的及意义 4 1.4 论文的主要内容 5 2 文本聚类 6 2.1