Understanding WeChat User Preferences and "Wow" Diffusion

WeChat is the largest social instant messaging platform in China, with 1.1 billion monthly active users. "Top Stories" a novel friend-enhanced recommendation engine WeChat, which users can read articles based on preferences of both their own and friends. Specifically, when user reads an article by opening it, "click" behavior private. Moreover, if clicks "wow" button, (only) her/his direct connections will be aware this action/preference. Based unique data, we aim to understand diffusion Top Stories at different levels. We have made some interesting discoveries. For instance, probability one negatively correlated number connected components that are formed friends, but click opposite. further study what extent users' predicted from connections. To address problem, present hierarchical graph representation learning model DiffuseGNN, capable capturing structure-based observations discovered above. Our experiments show proposed method significantly improve prediction performance compared alternative methods.