Balanced influence maximization in social networks based on deep reinforcement learning.

Area of Science:

• Social Network Analysis
• Information Propagation Dynamics
• Computational Social Science

Background:

• Balanced influence maximization is crucial for mitigating filter bubbles and echo chambers in social networks.
• Existing methods overlook entity correlations and require extensive diffusion sampling, limiting scalability.

Purpose of the Study:

• To propose a novel framework for balanced influence maximization that accounts for entity correlations and improves efficiency.
• To enhance the accuracy and scalability of balanced influence maximization in large social networks.

Main Methods:

• Developed a Balanced Influence Maximization framework based on Deep Reinforcement Learning (BIM-DRL).
• Introduced an entity correlation evaluation module using historical user behavior sequences.
• Designed a deep reinforcement learning-based seed node selection module for optimizing balanced influence.

Main Results:

• The proposed BIM-DRL framework effectively evaluates the impact of entity correlations on information propagation.
• BIM-DRL significantly improves balanced influence spread and balanced propagation accuracy compared to state-of-the-art methods.
• The framework demonstrates superior performance across six real-life network datasets.

Conclusions:

• BIM-DRL offers a more accurate and scalable solution for balanced influence maximization in social networks.
• The incorporation of entity correlations and deep reinforcement learning advances the field of influence maximization.
• This approach provides a robust method for promoting diverse information exposure and preventing echo chambers.