摘要： 为提高现有背景离群点检测算法背景子图划分的准确性，提出一种基于K-way谱聚类的背景离群点检测算法。构造图模型，对其进行K-way划分，使得到的背景子图具有解释性意义，从划分后的背景子图中获得离群点。实验结果表明，该算法的H指标提高50%，VI指标降低70%，其精确度有较大提高，且没有对图的结构进行改变，不会丢失重要信息。

关键词: K-way谱聚类, 二分法, 背景离群点, 随机游走, 背景子图, 图划分因子

Abstract: In order to improve the background subgraph classification accuracy of existing background outlier detection algorithm, this paper proposes a background outlier detection algorithm based on K-way spectral clustering. This paper establishes the diagram model, does the K-way partition to make it have explanatory significance for background subgraph, and gets the outliers from the background subgraph. Experimental results show that the accuracy of this algorithm is improved by 50% at H index and is reduced by 70% at VI index. There is no change with the structure of graph. So it cannot produce the problem of losting important information.

Key words: K-way spectral clustering, dichotomy, background outlier, random walk, background subgraph, graph partition factor

中图分类号: 

• TP391