Abstract:
The main challenge in viral marketing, that is powered by social networks, is to minimize
the seed set that will initiate the diffusion process and maximize the total influence
at its termination. The aim of this thesis is to study influence propagation models
under the influence maximization problem and to investigate the effectiveness of a
new model that is based on a multi-objective approach. We propose a Depth-Based
Diminishing Influence model (DBDM) that is based on adding nodes to the seed set
by considering influenced in-neighbors and how far these in-neighbors are from the
initial activated set. As an enhancement to our approach, we used a clustering mechanism
to help increase the influence spread. Several experiments were conducted to
compare between our approach and previous work. As a result, the selection of the
seed set under the DBDM model boosted the influence spread substantially compared
to previously proposed models.
