Abstract:

An approach to hidden group detection in social network based on hidden Markov evolution model is proposed.Different from conventional methods,we begin with reasonable assumptions for the micro-laws to determine whether at any given time a particular individual is in a community or not,based on which we are able to discover the individual dynamics that drive the evolution of the social groups in a community.Finally,we identify persistent groups over a time period long enough as potential hidden groups.Further analysis is made to ensure the high probability of these groups to be satisfactory results.Experiments on synthetic data as well as real communities (e.g. Enron email) are performed.

Key words: HMM;hidden group;probabilistic evolution