An Attention-based Model for Conversion Rate Prediction with Delayed Feedback via Post-click Calibration
Conversion rate (CVR) prediction is becoming in- creasingly important in the multi-billion dollar on- line display advertising industry. It has two ma- jor challenges: firstly, the scarce user history data is very complicated and non-linear; secondly, the time delay between the clicks and the correspond- ing conversions can be very large, e.g., ranging from seconds to weeks. Existing models usu- ally suffer from such scarce and delayed conver- sion behaviors. In this paper, we propose a novel deep learning framework to tackle the two chal- lenges. Specifically, we extract the pre-trained em- bedding from impressions/clicks to assist in con- version models and propose an inner/self-attention mechanism to capture the fine-grained personalized product purchase interests from the sequential click data. Besides, to overcome the time-delay issue, we calibrate the delay model by learning dynamic haz- ard function with the abundant post-click data more in line with the real distribution. Empirical experi- ments with real-world user behavior data prove the effectiveness of the proposed method.
PDF Abstract