SUBREDDIT Communities' Dynamic and Conflict Propagation

Community Interaction and Conflict on the Web present one of the most important threat facing the social Network users. In this work, we propose a new prediction model of future conflict propagation able to describe the social behavior of conflicts using information cascade trees. To achieve this objective, we use topic modeling, prediction models, and post-wise clustering to showcase predictability. Community clustering is incorporated to demonstrate how similar communities react and share information. Using Latent Dirichlet Allocation, posts are categorized into topics, revealing consistent language patterns in negative conflict posts. Machine learning techniques like random forest classifiers and linear models confirm that post properties can predict positivity or negativity. Clustering solely based on text demonstrates the distinguishability of positive and negative posts. This multi-faceted approach offers insights into understanding the complexities of the Reddit social network.