Community Feature Selection for Anomaly Detection in Attributed Graphs

详细信息    查看全文

• 关键词：Anomaly detection ; Feature selection ; Attributed graphs
• 刊名：Lecture Notes in Computer Science
• 出版年：2017
• 出版时间：2017
• 年：2017
• 卷：10125
• 期：1
• 页码：109-116
• 丛书名：Progress in Pattern Recognition, Image Analysis, Computer Vision, and Applications
• ISBN：978-3-319-52277-7
• 卷排序：10125

文摘

Anomaly detection on attributed graphs can be used to detect telecommunication fraud, money laundering, intrusions in computer networks, atypical gene associations, or people with strange behavior in social networks. In many of these application domains, the number of attributes of each instance is high and the curse of dimensionality negatively affects the accuracy of anomaly detection algorithms. Many of these networks have a community structure, where the elements in each community are more related among them than with the elements outside. In this paper, an adaptive method to detect anomalies using the most relevant attributes for each community is proposed. Furthermore, a comparison among our proposal and other state-of-the-art algorithms is provided.