Tracking individual behaviors in networks: An experimental demonstration

Tracking individual behaviors based on observations made from vast personal interaction network has become a major concern and interest for the policing community as well as for the business/commercial players. While the policing community resort to personal networks in order to predict and prevent adverse events, the commercial players want to track opportunities for online advertisement, market identification, personalized product suggestions etc. Recently, revolutionary advances in digital media technology have enabled one to collect, store and analyze massive amounts of personal networking data. Unlike traditional tracking problems, the observations and the inferred targets are highly irregular in nature; they do not evolve or be observed according to established mathematical models. In this paper, we demonstrate an experimental approach for tracking hidden qualities of individuals by observing their closer connections in the personal networks they belong to. We model the hidden features of individuals through hidden Markov random fields (HMRF) and propose a modified observation model in order to simplify the tracking algorithm. We test our algorithm on a fictitious scale-free personal network dataset and report high accuracy through objective performance metrics.