A multistrain bacterial model for link prediction

In this paper we introduce a novel model for link prediction in social network based on a quantitative growth and diffusion model of node features which are used to compute candidate links ranking. The model is inspired by the biological mechanisms which regulates bacteria reproduction and their transfer among subjects through physical contact. The basic idea is that nodes infect their neighborhood with their own bacteria strains, i.e. node identifiers, and the infections are iteratively propagated on the network over the time. The value of the mutual strains of infection in a pair of nodes is then used for ranking the potential arc joining the nodes. The iterative process of growth-infection and the mutual link ranking computation has been implemented and tested on widely accepted social network datasets. Experiments shows that the proposed model outperform state of the art ranking algorithms.