Abstract

Event detection and forecasting in social media networks, such as disease outbreak and air pollution event detection, have been formulated as an anomalous connected subgraph detection problem. However, the huge search space and the sparsity of anomaly events make it difficult to solve this problem effectively and efficiently. This paper presents a general framework, namely anomalous connected subgraph scanning (GraphScan) which optimizes a large class of sophisticated nonlinear nonparametric scan statistic functions, to solve this problem in attributed social media networks. We first transform the sophisticated nonlinear nonparametric scan statistics functions into the Price-Collecting Steiner Tree (PCST) problem with provable guarantees for evaluating the significance of connected subgraphs to indicate the ongoing or forthcoming events. Then, we use tree decomposition technique to divide the whole graph into a set of smaller subgraph bags, and arrange them into a tree structure, through which we can reduce the search space dramatically. Finally, we propose an efficient approximation algorithm to solve the problem of anomalous subgraph detection using the tree of bags. With two real-world datasets from different domains, we conduct extensive experimental evaluations to demonstrate the effectiveness and efficiency of the proposed approach.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call