STUDY ON OPINION LEADER IN ONLINE SOCIAL NETWORK USING COMPUTATIONAL INTELLIGENCE

Abstract

Context: In the current scenario, the online social network has plenty of perspectives to interact with the other person, share information and ideas, and discover and realize the new thing within a single click. Social networking sites provide web-enabled resources that make human life more convenient and contented through the more comfortable communication method. Whenever people find any decisive situation in their daily lives, they share their viewpoints and opinions through blogs, social networking services, reviews, pictures, videos, etc. In the social network, an opinion leader is a person who has the ability to deflect human decision-making through their dexterity, knowledge, experience, and attitude. Nowadays, organizations appoint opinion leaders to promote their products as part of marketing strategies. The applicability of opinion leaders is very abundant in real-world applications like marketing, finance, recommender system, healthcare, consumer behavior, online learning and knowledge communities, blogosphere, and many more diverse fields. Therefore, this study represents an organized, systematic, and arranged effort that determines the identification, power, and applicability of opinion leaders in the online social network. Objective: The objective of the entire study has been classified into three segments.  The primary objective of the study is to discover the optimal opinion leaders in the online social network based on computational intelligence techniques.  The second objective focuses on presenting the significance and power of opinion leaders for the diffusion of products in the online social network.  The third objective is exploring the applicability of the opinion leader through the online social network in healthcare.

vi Methodology: For achieving the mentioned objectives, this study utilizes computational intelligence techniques like nature-inspired metaheuristic algorithms, game theory, graph neural networks, and fuzzy logic due to the tremendous applicability to solve natural world problems. Following strategies are used to achieve the targeted objectives:  For achieving the first objective, innovative and novel computational intelligence techniques are implemented to identify the suitable opinion leader in the online social network. The social network-based variant of two nature-inspired metaheuristic algorithms, the firefly algorithm and whale optimization algorithm, respectively, are used to find the opinion leaders.  To accomplish the second objective, two approaches; Game theory and Graph Neural Network-based, have addressed the importance and power of the opinion leader for the diffusion of products in the online social network. The game theory-based strategy is used to elucidate the coalition of opinion leaders, while Graph Neural Network-based technique proposed a reputation and trust-based unique model to show the relevance of opinion leaders for information diffusion.  To attain the third objective, the relevance and applicability of opinion leaders are explored in healthcare. In the ongoing Covid-19 pandemic course, people spread various COVID-19 related rumors and hoaxes on social networks, which incredibly negatively influences civilization. A reputation-based opinion leader identification algorithm is designed that identifies opinion leaders to control the spreading of COVID-19 rumors. Results: The outcomes of the study are as follows:  A social network-based firefly algorithm is designed to identify the opinion leader within the local communities and globally. The Accuracy, Precision, Recall, and F1- score of the firefly-based model is 0.94, 0.93, 0.95, and 0.94, respectively.  A modified Louvain community partitioning algorithm has been designed to identify the communities in the network. The algorithm’s modularity gain and running time are around 14% lesser than the original Louvain method.

vii  A social network-based whale optimization algorithm is addressed that suitably recognizes the opinion leaders based on various optimization functions. The Accuracy, Precision, Recall, and F1-score of the model are 0.95, 0.94, 0.95, and 0.95, respectively, based on 12 standard benchmark functions.  An innovative community partitioning algorithm is also designed to find the communities based on neighborhood similarity. The total running time of the algorithm is 11% faster than the other standard algorithm. Also, the other parameters like Node attribute similarity, Common neighbor similarity, Average community density, etc., are reduced to around 12%.  A new Game theory-based Opinion Leader Detection algorithm is presented to identify the coalitions of opinion leaders with the maximum synergy, marginal payoff, and Shapley value. The Accuracy, Precision, Recall, and F1-score of the model are around 0.94, 0.94, 0.95, and 0.95, respectively. Also, the diffusion rate is enhanced by approximately 15% over other SNA measures.  An exclusive Graph Neural Network (GNN) for Opinion Leader Identification (GOLI) model is proposed that utilizes the power of GNN to categorize the opinion leaders and their impact on the diffusion of products in the online social network. The GOLI model obtained around 91% training accuracy and 92% testing accuracy with an approximately 1% error rate. The Accuracy, Precision, Recall, and F1-score of the model are around 0.95, 0.96, 0.96, and 0.96, respectively.  An Opinion Leader-based Rumor Detection (OLRD) algorithm is designed to show opinion leaders’ applicability and significance for controlling the COVID-19 rumors. The proposed approach reduced the total number of diffusers by 26% faster, spread veracity around 22% more quickly, and impacted approximately 23% faster than other SNA measures.  A Reputation-based Opinion Leader Identification (ROLI) algorithm is defined to find the opinion leaders in the online social network. The proposed model produces 91% Accuracy,93% Precision, 95% Recall, and 94% F1-score.