基于密度半径自适应选择的K-均值聚类算法

K-均值算法聚类速度快,易于实现,且对数据依赖度低,在文本聚类中得到广泛应用.然而,由于聚类初始中心点选择的随机性,传统K-均值算法以及其变种的聚类结果会产生较大的波动.文章对K-均值算法进行了改进,通过自适应选择最佳密度半径进而优化聚类初始中心选择的方法,得到一种适合文本数据聚类分析的改进算法.实验表明,该算法能够生成质量较高而且波动性较小的聚类结果.

K-means Clustering Algorithm Based on Self-Adoptively Selected Density Radius

K-means is one of the widely used text clustering techniques due to its rapidity, simplicity and high scalability. However, owing to random selection of initial centers, unstable results were often obtained while using traditional K -means and its variants. A technique of optimizing initial centers of clustering is proposed based on self-adoptively selecting density radius. The experimental result shows that the K -means with the proposed technique can produce cluster results with high purity as well as good stableness.