A resilience control method for mitigating the sudden change in online group opinion based on Q-Learning and PSO

Abstract

In today's era, individuals can speak and communicate online anytime and anywhere. The Online Group Opinion (OGO), formed by the online statements of netizens, is regarded as a complex system. OGO is easy to get rapid and unpredictable sudden changes under external stimuli such as social emergencies, which may further lead to online violence, catalyze spread of emergencies, posing significant threats to social organization and cybersecurity. This paper proposes an innovative method to mitigate sudden changes in OGO using Q-learning from Reinforcement Learning and the Particle Swarm Optimization (PSO) algorithm, aiming to mitigate these sudden changes by controlling the resilience of the OGO. First, an OGO resilience index based on catastrophe theory and resilience theory is established to portray OGO sudden changes phenomena. Next, a resilience control algorithm integrated with Q-learning to optimize particles’ parameter update process is proposed, called QLPSOND. QLPSOND is used to control the system’s resilience to improve its ability to resist external stimuli, thus mitigating the sudden changes crisis. Through comparisons with other three algorithms on two online forums datasets, our proposed QLPSOND resilience control method demonstrates significant advantages in addressing complex social control issues, with average fitness improvement of 37.59 % and 7.98 %, and best fitness improvement of 31.66 % and 10.28 %, respectively, compared to most competitive state-of-the-art baseline. Additionally, this paper further discusses the practical significance of resilience control strategies and management implications. These efforts can extend the opinion dynamics approaches in online environments and assist enterprises and governments in monitoring and managing online communities.