Influence Maximization on Competitive and Complementary Independent Cascade Model Based on Rating-Aware Bidirectional Gated Recurrent Unit

Abstract

Influence maximization aims to maximize the dissemination of products, ideas, and other entities by selecting seed nodes in social networks. Although significant progress has been made in single-entity propagation and competitive entity scenarios, the mechanisms of influence propagation in social networks where complementary and competitive entities coexist remain underexplored, limiting the applicability of existing models in real-world settings. To address this, this paper proposes an innovative influence maximization framework based on user historical behavior, incorporating the Competitive and Complementary Independent Cascade (CCIC) model and the User Historical Behavior-based Cost-Effective Lazy Forward (UHBCELF) algorithm. First, the CCIC model is introduced to systematically describe the process of propagating a target entity in a social network with multiple related entities. Second, to capture the relationships between entities, the rating-aware bidirectional gated recurrent unit is designed, which predicts the probability of users adopting the target entity by mining implicit patterns in their historical behavior. Finally, the UHBCELF algorithm is developed, which optimizes seed node selection by leveraging user historical behavior data, significantly enhancing propagation effectiveness. Experimental results demonstrate that the proposed method outperforms state-of-the-art methods on four public datasets, achieving improvements of 6.8% and 7.4% in key metrics such as F1-score and expected spread range, respectively. This work provides a novel perspective and practical approach for addressing influence maximization in multi-entity social networks, offering valuable insights for real-world applications.

Keywords: Influence Maximization, Historical Behavior Sequence, Gated recurrent unit, Social Network