Deep Position-wise Interaction Network for CTR Prediction

Abstract

Click-through rate(CTR) prediction plays an important role in online advertising and recommender systems. In practice, the training of CTR models depends on click data which is intrinsically biased towards higher positions since higher position has higher CTR by nature. Existing methods such as actual position training with fixed position inference and inverse propensity weighted training with no position inference alleviate the bias problem to some extend. However, the different treatment of position information between training and inference will inevitably lead to inconsistency and sub-optimal online performance. Meanwhile, the basic assumption of these methods, i.e., the click probability is the product of examination probability and relevance probability, is oversimplified and insufficient to model the rich interaction between position and other information. In this paper, we propose a Deep Position-wise Interaction Network (DPIN) to efficiently combine all candidate items and positions for estimating CTR at each position, achieving consistency between offline and online as well as modeling the deep non-linear interaction among position, user, context and item under the limit of serving performance. Following our new treatment to the position bias in CTR prediction, we propose a new evaluation metrics named PAUC (position-wise AUC) that is suitable for measuring the ranking quality at a given position. Through extensive experiments on a real world dataset, we show empirically that our method is both effective and efficient in solving position bias problem. We have also deployed our method in production and observed statistically significant improvement over a highly optimized baseline in a rigorous A/B test.

Click-Conversion Multi-Task Model with Position Bias Mitigation for Sponsored Search in eCommerce

“…𝑤 𝑐𝑣𝑟 𝑖 can be calculated similarly. In addition to the weighted MRR, we also apply the widely-used evaluation metrics: standard MRR, AUC, and position-wise AUC (PAUC) [11] to evaluate model performance.…”

“…• We propose to jointly learn position-bias-free CTR and CVR prediction models in a multi-task learning framework. By mitigating position bias, the proposed models achieve comparable performance as state-of-the-art models on CTR prediction and significant performance improvement on CVR prediction regarding weighted Mean-Reciprocal-Rank (MRR) [19], MRR, position-wise AUC (PAUC) [11], and AUC. • We conduct experiments on real-world E-commerce sponsored product searches.…”

Click-through rate prediction in online advertising: A literature review

“…Among these characteristics, advertising position is significantly associated with CTR in sponsored search advertising (Yang et al, 2018). In this sense, the position information (Yuan et al, 2020;Huang et al, 2021) and the externality effect from surrounding advertisements (Deng et al, 2018) are definitely useful to improve prediction results. In this direction, it calls for more attention on the power of advertising characteristics in CTR prediction.…”

NMA: Neural Multi-slot Auctions with Externalities for Online Advertising

“…However, optimization problem of the choice model based on 0-1 integer program is NP-hard and its assumption of user behaviors independent of the ad ranking order is inconsistent with our intuition [21]. Deep Position-wise Interaction Network (DPIN) [12] was modeled in Meituan for position-dependent externality, which predicts the CTR of each ad in each position and then split the multiple slots allocation into multiple rounds of single slot auction based GSP. DPIN does not consider ad-dependent externalities and is not truth-telling.…”