Multistage Competitive Opinion Maximization With Q-Learning-Based Method in Social Networks.

Area of Science:

• Social Network Analysis
• Computational Social Science
• Artificial Intelligence

Background:

• Competitive Opinion Maximization (COM) is an emerging field focused on strategic opinion propagation in social networks.
• Existing COM research is limited, primarily addressing known competitor strategies and static user opinions.
• There's a need for advanced COM frameworks that accommodate dynamic user opinions and unknown competitor behaviors.

Purpose of the Study:

• To propose a novel framework for multistage Competitive Opinion Maximization (COM).
• To address limitations in current COM approaches, specifically dynamic opinion changes and unknown competitor strategies.
• To maximize relative effective opinions in complex social network environments.

Main Methods:

• Developed a Q-learning-based opinion maximization framework (QOMF) with dynamic opinion propagation and a seeding process.
• Designed a realistic opinion propagation model integrating activation and dynamic opinion processes, proving its convergence.
• Implemented a multistage Q-learning seeding scheme to handle both known and unknown competitor strategies.

Main Results:

• The proposed QOMF framework demonstrates superior performance compared to existing benchmarks.
• The dynamic opinion propagation model achieves convergence, ensuring realistic opinion spread simulation.
• The Q-learning seeding scheme effectively identifies optimal seed nodes under varying competitor strategies.

Conclusions:

• The QOMF framework offers a significant advancement in Competitive Opinion Maximization (COM).
• The approach successfully addresses dynamic user opinions and unknown competitor strategies in social networks.
• Experimental validation confirms the framework's effectiveness in maximizing relative effective opinions.