Learning Attribute-weighted Voter Model over Social Networks


Yuki Yamagishi, Kazumi Saito, Kouzou Ohara, Masahiro Kimura, Hiroshi Motoda ;
Proceedings of the Asian Conference on Machine Learning, PMLR 20:263-280, 2011.


We propose an opinion formation model, an extension of the voter model that incorporates the strength of each node, which is modeled as a function of the node attributes. Then, we address the problem of estimating parameter values for these attributes that appear in the function from the observed opinion formation data and solve this by maximizing the likelihood using an iterative parameter value updating algorithm, which is efficient and is guaranteed to converge. We show that the proposed algorithm can correctly learn the dependency in our experiments on four real world networks for which we used the assumed attribute dependency. We further show that the influence degree of each node based on the extended voter model is substantially different from that obtained assuming a uniform strength (a naive model for which the influence degree is known to be proportional to the node degree), and is more sensitive to the node strength than the node degree even for a moderate value of the node strength.

Related Material