Robust Representations for Response Prediction

Measuring the performance of display advertising is an important problem in estimating campaign effectiveness and understanding user behaviors. The two key performance indicators are the click-through rates (CTR) of the ads and conversion rates (CVR) on the advertisers website. Existing approaches for conversion prediction and for click prediction usually look at the two problems in isolation. However there is considerable benefit in jointly solving the problems as the two goals are often intertwined. In this chapter, we aim to predict the conversion response of the users by jointly examining the past purchase behavior and the click response behavior. To achieve this, we explicitly model the temporal dynamics between the click response and purchase activity into a unified framework. More specifically, we propose a novel matrix factorization approach named the dynamic collective matrix factorization (DCMF) to address this problem. Our model considers temporal dynamics of post-click conversions and also takes advantages of the side information of users, advertisements, and items. An efficient optimization algorithm based on stochastic gradient descent is presented in the chapter. We further show that our approach can be used to measure the effectiveness of advertisements. Our approach and several representative baselines are evaluated on a public dataset and a real-world marketing dataset. Extensive experimental results demonstrate that our model achieves significant improvements over the baselines.