On Modeling Brand Preferences in Item Adoptions

In marketing and advertising, developing and managing brands value represent the core activities performed by companies. Successful brands attract buyers and adopters, which in turn increase the companies’ value. Given a set of user-item adoption data, can we infer brand effects from users adopting items? To answer this question, we develop the Brand Item Topic Model (BITM) that incorporates users’ brand preferences in the process of item adoption by the users. We evaluate our model using synthetic and two real world datasets against baseline models which do not consider brand effects. The results show that BITM can determine users who demonstrate brand preferences and predict item adoptions more accurately.