Real-Time Bidding User Response Prediction

Abstract

With the advent of the "digital age," online advertising is changing the way in which advertisers reach out to and interact with potential customers. Driven by the trend of increased integration and development of technology and the harnessing of the growing volume of data available, online display advertising has seen significant change in recent years through the rise of real-time bidding (RTB), which has seen significant growth since its emergence in 2009. My goal is to explore and utilize a real-world RTB bidding dataset to approach the task of demand-side problem of user response prediction in the RTB setting. In order to do so, I will exploit contextual features to create advertiser-specific user response models, which, when combined with finely tuned market price forecasts, can serve as input for determining the optimal bid for an ad impression. Finally, evaluation of policies will be conducted using a test set of the original RTB dataset.