Based on Stochastic Information Diffusion, Link Prediction

PROJECT TITLE :

Link Prediction Based on Stochastic Information Diffusion

ABSTRACT:

Link prediction (LP) in networks aims to determine future interactions among elements; it is an essential piece of Machine Learning software in a variety of fields, ranging from genomics to social networks to marketing, particularly in e-commerce recommender systems. LP is an abbreviation for "link prediction" in networks. Although a great number of LP techniques have been developed in the prior art, the vast majority of them take into account only the static structures of the underlying networks and very rarely incorporate the information flow of the network. Exploiting the impact of dynamic streams, like information diffusion, is still a question that needs to be answered by LP researchers. The ability of nodes to receive information from outside of their social circles as a result of information diffusion, which in turn can have an effect on the formation of new links. In this study, we investigate the effects of LP by utilizing two distinct diffusion methodologies: susceptible-infected-recovered and independent cascade. As a consequence of this, we suggest the progressive-diffusion (PD) method for LP, which is based on the dynamics of the propagation of the nodes. The model that is being proposed makes use of a stochastic discrete-time rumor model that is centered on the propagation dynamics of each node. It has a small footprint in terms of both memory and processing power, and it lends itself well to both parallel and distributed processing implementations. In conclusion, we also present a new metric for the evaluation of LP methods that takes into account both the information diffusion capacity and the LP accuracy. The effectiveness of the proposed method in comparison to the prior art in both criteria is attested to by experimental results on a series of benchmarks that were used as comparisons.

Did you like this research project?

To get this research project Guidelines, Training and Code... Click Here