Comparative Influence Maximization With Adaptive Policy

Influence maximization in social networks has received much research attention in recent years. It asks to select a set of influential users as seed nodes, so as to trigger the maximum influence spread across the network. However, existing works mainly focus on single item or a single-round propagation model, which seriously does not match the complex real-world scenario. In this paper, we propose the Multi-Round Comparative Independent Cascade model (MRCom-IC), which accurately models the simultaneous propagation of multiple items over multiple rounds in the social network. Two natural optimization problems then arise: the multi-round self-influence maximization problem and the multi-round complementary influence maximization problem. Both problems are classified as non-deterministic polynomial-time hard (NP-hard). We thoroughly study the two problems under both non-adaptive and adaptive scenarios, and propose efficient and effective approximation algorithms based on the reverse reachable set (RR-set) technique and the “sandwich approximation” strategy. Extensive experimental results show that the proposed algorithms significantly outperform the comparison methods in various aspects.