Multilevel learning based modeling for link prediction and users’ consumption preference in Online Social Networks

摘要: Abstract The problem with predicting links in Online Social Networks (OSNs) is having to estimate the value of a link that can represent relationship between social media users. evolution OSN influenced by structure network and interaction preferential behaviors users have long converged sociologists. However, conventional methods treat these isolation. Therefore, roles users’ historical preferences dynamic are still not clear as how things affect OSN. Link prediction for new who created or small fundamental OSNs. To start creating networks such users, be used recommend friends user consumption preferences. In this paper, we propose novel direct latent models user’s an platform. We also introduce multilevel deep belief learning-based model achieve high accuracy. evaluate performance our model, elaborated several measures datasets from Facebook, Amazon Google+ validate result evaluation shows proposed provides significantly improved over other methods.