基于中心距序降维的聚类算法

Clustering Algorithm Based on Dimension Reduction by Center Distance Order

为提高金融业务数据集上的聚类质量和聚类效率,提出簇的直径、簇间的相似度这2个概念。利用距离尺度降维的中心距序降维法,将多维数据降至一维,在一维上利用自适应排序聚类算法ASC聚类。该算法和传统的Cobweb算法、K-means算法做对比,实验表明该方法能提高簇间相似度,最大提高200%。

Aiming to improve the clustering quality and efficiency on banking services datasets, this paper proposes the concepts of cluster diameter and the similarity measurement between clusters. It modifies multi-dimensional data to one dimension by dimension reduction based on distance order. It clusters the one dimension data with a self-Adaptive Sort Clustering（ASC） algorithm. This paper conducts extensive experiments to show that this algorithm can improve the cluster similarity and reduce the clustering time compared with Cobweb and K-means algorithms. The cluster similarity can be approximately improved by 200%.