|
Abstract:
|
Considerable work has been done to address the topological features to predict the community structure, dynamic relationships, and the evolution pattern of dynamic networks. In particular, It is important to understand the excessive communications between nodes within a social network because such excessive activities provide an insight into the pattern of communication between nodes. In some cases, these excessive communications would be an indicator of fraudulent activities. In the present work, we implement a new approach to study the joint dynamics of multiple nodes by developing multivariate time series models to identify the dynamic relationship of a critical ego with other egos in the neighborhood ego networks using multivariate time series models. Furthermore, in social networks, the detection of anomalous topic activities associated with the influential nodes is an unsolved problem. Topics across several documents can be derived using the probabilistic model, Latent Dirichlet’s allocation (LDA). We discuss here a new approach by combining the LDA and the scan statistics to investigate whether there exists a statistically significant anomalous topic activity.
|
