Abstract: In order to detect the boundary points of clusters effectively, the concept of grid-entropy and a grid-entropy-based boundary points detecting algorithm Greb were proposed, which detected boundary points by the value of grid-entropy and only needed to scan the datasets twice. As shown in the experimental results, Greb can detect boundary points effectively and efficiently on various datasets.

Key words: boundary points, grid-entropy, clusters

摘要： 为了快速有效地检测聚类的边界点，提出了网格熵的概念和基于网格熵的边界点检测算法Greb。该算法利用网格熵的大小来判定聚类的边界点，且只对数据集进行两遍扫描。实验结果表明，对含有任意形状、不同大小以及不同密度且带有噪声的数据集，该算法能快速有效地检测出聚类的边界点。

关键词: 边界点, 网格熵, 聚类