We study the problem of maximizing the expected spread of influence within a social network. In this report, we used the Twitter accounts of "ProudBoys" members and of other users associated with them which form of snapshot a social network in the year 2020. We analyzed the influence maximization problem in several of the most widely studied models in social network analysis such as Independent Cascade, Decreasing Cascade, Weighted Cascade, TRIVALENCY, Linear Threshold, and Generalized Threshold models, along with 3 influence maximization algorithms: Naive Greedy, Cost Effective Lazy Forward (CELF) and Maximum Influence Arborescence (MIA). We compared the results from each algorithm paired with each model. As result, we detected accounts with the most influence on the network as it pertains to diffusion of information.
The widespread use of the internet has led to billions of people being connected through online social media platforms like TikTok, Twitter, and Instagram. These platforms generate a large amount of data, which has led to increased research on social networks. In addition to being a means of communication, social networks also serve as a platform for sharing information, providing public services, and marketing.
In recent years, with the popularity of social networks, the influence maximization problem has become a pressing issue in this field. The Influence Maximization problem identifies a small subset of the most important influencers in the network to tackle some real-world problems and activities(\cite{SINGH20227570}). Numerous techniques to improve the performance of Influence Maximization have been proposed. In this project, first, we will explore how to model the diffusion process to propagate the information by adapting several well-accepted diffusion models. Second, we will compare and analyze the outcome of existing Influence Maximization algorithms deployed on our dataset. Our goal is to identify the top influential users in Twitter's network and determine which models and algorithms perform best on the dataset.
