Self-Supervised Hypergraph Representation Learning for Sociological Analysis

Modern sociology has profoundly uncovered many convincing social criteria for behavioral analysis. Unfortunately, of them are too subjective to be measured and very challenging presented in online networks (OSNs) the large data volume complicated environments explored. On other hand, mining techniques can better find patterns but leave behind unnatural understanding humans. Although there some works trying integrate observations specific tasks, they still hard applied more general cases. In this paper, we propose a fundamental methodology support further fusion sociological criteria. Our highlights three-fold: First, an effective hypergraph awareness fast line graph construction framework. The indicate interactions between individuals their because each edge (a.k.a hyperedge) contains than two nodes, which is perfect describe social. A treats environment as super node with underlying influence different environments. way, go beyond traditional pair-wise relations explore richer under various criteria; Second, novel hypergraph-based neural network learn flowing from users users, environments, learned via task-free method, making our model flexible tasks analysis; Third, both qualitative quantitive solutions effectively evaluate most common like conformity, equivalence, environmental evolving polarization. extensive experiments show that framework user behaviors