AE-UPCP: Seeking Potential Membership Users by Audience Expansion Combining User Preference with Consumption Pattern

Abstract

Many online video websites or platforms provide membership service, such as YouTube, Netflix and iQIYI. Identifying potential membership users and giving timely marketing activities can promote membership conversion and improve website revenue. Audience expansion is a viable way, where existing membership users are treated as seed users, and users similar to seed users are expanded as potential memberships. However, existing methods have limitations in measuring user similarity only according to user preference, and do not take into account consumption pattern which refers to aspects that users focus on when purchasing membership service. So we propose an Audience Expansion method combining User Preference and Consumption Pattern (AE-UPCP) for seeking potential membership users. An autoencoder is designed to extract user personalized preference and CNN is used to learn consumption pattern. We utilize attention mechanism and propose a fusing unit to combine user preference with consumption pattern to calculate user similarity realizing audience expansion of membership users. We conduct extensive study on real datasets demonstrating the advantages of our proposed model.