Efficient methods to select Top-K propagators based on distance and radius neighbor

Abstract

The problem of influence maximization aims to select Top-K propagators according to the available budget in which when these propagators are selected the influence coverage is maximized. While most existing approaches aim to increase the influence, spread based on centrality measures only, without considering the distance in which each selected seed should be far away from another seed and hence can reduce significantly the influence spread performance. This observation motivated us to design two new algorithms that are based on a new proposed metrics namely "Radius-Neighborhood Degree" and "Radius-Weighted Edges". The proposed two new algorithms are for selecting Top-K propagators named Radius-neighborhood degree over d-hops "RND d-hops" algorithm and combined radius-propagation probability threshold over d-hops and "RND d-hops" named "CPRND d-hops" algorithm. The "RND d-hops" algorithm selects Top-K propagators over d-hops, based on "Radius-Neighborhood Degree" metric that counts the degree of nodes till graph radius. The "CPRND d-hops" algorithm selects Top-K propagators to improve the selection of "RND d-hops" algorithm of Top-K propagators. The "CPRND d-hops" algorithm is based on "Radius-Weighted Edges" metric, that counts only the nodes edges which satisfy a certain propagation threshold considering "RND d-hops" algorithm selection. We conducted various experiments on large-scale graphs and compared them to the existing state of art approaches. The experimental results demonstrated that the proposed algorithms outperform the existing algorithms in term of influence achieved.