Comprehensive Audience Expansion based on End-to-End Neural Prediction

In current online advertising applications, look-alike methods arevaluable and commonly used to identify new potential users, tack-ling the difficulties of audience expansion. However, the demo-graphic information and a variety of user behavior logs are highdimensional,noisy, and increasingly complex, which are challengingto extract suitable user profiles. Usually, rule-based and similarity-based approaches are proposed to profile the users’ interests andexpand the audience. However, they are specific and limited inmore complex scenarios.In this paper, we propose a new end-to-end solution, unifyingthe feature extraction and profile prediction stages. Specifically,we present a neural prediction framework and leverage it with theintuitive audience feature extraction stages. We conduct extensivestudy on a real and large advertisement dataset. The results demon-strate the advantage of the proposed approach, not only in accuracybut also generality.