Influence-Based Community Partition With Sandwich Method for Social Networks

Community partition is an important problem in many areas, such as biology networks and social networks. The objective of this to analyze the relationships among data via network topology. In article, we consider community under independent cascade (IC) model We formulate a combinatorial optimization that aims at partitioning given into disjoint $m$ communities. maximize sum influence propagation through maximizing it within each community. existing work shows maximization for (IMCPP) NP-hard. first prove function IMCPP IC neither submodular nor supermodular. Then, both supermodular upper bound lower are constructed proved so sandwich framework can be applied. A continuous greedy algorithm discrete implementation devised problems. two problems gets notation="LaTeX">$1-1/e$ approximation ratio. also present simple solve original apply guarantee data-dependent factor. Finally, our algorithms evaluated on three real datasets, which clearly verifies effectiveness method problem, well advantage against other methods.