Deep Learning for Online Display Advertising User Clicks and Interests Prediction

Abstract

In this paper, we propose a deep learning based framework for user interest modeling and click prediction. Our goal is to accurately predict (1) the probability that a user clicks on an ad, and (2) the probability that a user clicks a specify type of campaign ad. To achieve the goal, we collect page information displayed to users as a temporal sequence, and use long-term-short-term memory (LSTM) network to learn latent features representing user interests. Experiments and comparisons on real-world data shows that, compared to existing static set based approaches, considering sequences and temporal variance of user requests results in an improvement in performance ad click prediction and campaign specific ad click prediction.