User-driven competitive influence maximization in social networks

Abstract

Online social networks have emerged as pivotal platforms where users not only interact but also influence each other's decisions and preferences. As these networks grow in complexity, understanding and leveraging influence dynamics within networks have become essential, particularly for businesses and marketers. Competitive Influence Maximization (CIM) in online social networks has garnered significant interest, focusing on maximizing influence spread among multiple entities. However, recent research on CIM often overlooks the differences in user preferences, which realistically impact the propagation of competitive influence. To address this issue, we introduce the User-Driven Competitive Linear Threshold (UDCLT) model. This model takes into account user preference differences for two distinct brands within the identical product category, thereby formulating the User-Driven Competitive Influence Maximization (UDCIM) problem. Based on community structure, we introduce a novel measure, namely Topology Importance (TI), to assess a node's potential influence within a social network by considering its connections within and across communities. To resolve the UDCIM problem effectively, we develop a novel two-phase algorithm, the Community-based Dual Influence Assessment (CDIA) algorithm, which integrates Topology Importance and Dual Influence to identify seed nodes. Various experiments are conducted on four real-world datasets, illustrating the efficiency and effectiveness of the CDIA algorithm in addressing the UDCIM problem.