Complex Network Coarse Graining And Anomaly Detection

Anomaly detection is an important problem in the study of complex network theory, which has attracted attention of many researchers from different specialized fields in recent years. Originally, techniques focused on anomaly detection in static graphs. Actually, many real networks are changing over time, the structure and attribute of that has close connection with time, this kind of networks contains properties of time is known as temporal network. In the study of theory and practical application, taking time attribute into anomaly detection of complex network will be greater significance.Many researchers have studied anomaly detection of temporal network at different levels, however no one take time scale or spatial scale into consideration up to now. Based on the above problems, this study chooses coarse-graining and anomaly detection on complex network as core study content, which detects anomaly in temporal networks and analyzes the anomaly in network under different time and spatial scale and correlation of anomalies.From perspective of time scale, this study introduces the method about how to coarsen temporal network based on time scale and proposes the method of detecting anomaly of nodes behavior in temporal network. This method focuses on studying the diversity and correlation of abnormal behavior of nodes under different time scales. The experimental results show that time scale has an important effect on anomaly detection about nodes behavior in temporal network. As a result, using reasonable time scale is important in detecting the abnormal behavior of nodesFrom perspective of spatial scale, this study introduces the method about how to coarsen network on spatial scale and proposes the method of detecting anomaly of static network nodes. This method focuses on analyzing the correlation of abnormal nodes between micro and macro spatial scale. The experimental results show that abnormal nodes under micro spatial scale will make a significant influence under the macro spatial scale.Due to time constraints, this work mainly studies the network anomaly and relationship of different anomalies unilaterally based on time scale or spatial scale. Considering combine time scale and spatial scale to study this problem in-depth is the next work.