Deep Spatio-Temporal Neural Networks for Click-Through Rate Prediction

Ouyang, Wentao; Zhang, Xiuwu; Li, Li; Zou, Heng; Xing, Xin; Liu, Zhaojie; Du, Yanlong

doi:10.1145/3292500.3330655

Cited by 50 publications

(51 citation statements)

References 26 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The dataset has been widely used as benchmarks for CTR Prediction. Following other works [25,26], we use the ad logs from 2015-04-28 to 2015-05-18 for training, those on 2015-05-19 for validation, and those on 2015-05-20 for testing. Moreover, only users with more than one page are kept.…”

Section: Experiments Setup 51 Datasetsmentioning

confidence: 99%

Modeling Users' Contextualized Page-wise Feedback for Click-Through Rate Prediction in E-commerce Search

Fan

et al. 2022

Proceedings of the Fifteenth ACM International Conference on Web Search and Data Mining

View full text Add to dashboard Cite

Modeling user's historical feedback is essential for Click-Through Rate Prediction in personalized search and recommendation. Existing methods usually only model users' positive feedback information such as click sequences which neglects the context information of the feedback. In this paper, we propose a new perspective for context-aware users' behavior modeling by including the whole page-wisely exposed products and the corresponding feedback as contextualized page-wise feedback sequence. The intra-page context information and inter-page interest evolution can be captured to learn more specific user preference. We design a novel neural ranking model RACP(Recurrent Attention over Contextualized Page sequence), which utilizes page-context aware attention to model the intra-page context. A recurrent attention process is used to model the cross-page interest convergence evolution as denoising the interest in the previous pages. Experiments on public and real-world industrial datasets verify our model's effectiveness.

show abstract

Section: Experiments Setup 51 Datasetsmentioning

confidence: 99%

Modeling Users' Contextualized Page-wise Feedback for Click-Through Rate Prediction in E-commerce Search

Fan

et al. 2022

Proceedings of the Fifteenth ACM International Conference on Web Search and Data Mining

View full text Add to dashboard Cite

show abstract

“…They can be regarded as complement work with our approach. State-of-the-art methods have found the effectiveness of modeling users' historical behaviors for CTR prediction [8,10,16,17,26,27,31,40,41]. DIN [41] notices that a user may have multiple interests and uses attention mechanism to learn the representation of user interests from historical behaviors with respect to a certain candidate item.…”

Section: Related Work 21 Ctr Predictionmentioning

confidence: 99%

Deep Multifaceted Transformers for Multi-objective Ranking in Large-Scale E-commerce Recommender Systems

Ding

Wang

et al. 2020

Proceedings of the 29th ACM International Conference on Information &Amp; Knowledge Management

View full text Add to dashboard Cite

Recommender Systems have been playing essential roles in ecommerce portals. Existing recommendation algorithms usually learn the ranking scores of items by optimizing a single task (e.g., Click-through rate prediction) based on users' historical click sequences, but they generally pay few attention to simultaneously modeling users' multiple types of behaviors or jointly optimize multiple objectives (e.g., both Click-through rate and Conversion rate), which are both vital for e-commerce sites. In this paper, we argue that it is crucial to formulate users' different interests based on multiple types of behaviors and perform multi-task learning for significant improvement in multiple objectives simultaneously. We propose Deep Multifaceted Transformers (DMT), a novel framework that can model users' multiple types of behavior sequences simultaneously with multiple Transformers. It utilizes Multi-gate Mixture-of-Experts to optimize multiple objectives. Besides, it exploits unbiased learning to reduce the selection bias in the training data. Experiments on JD real production dataset demonstrate the effectiveness of DMT, which significantly outperforms state-ofart methods. DMT has been successfully deployed to serve the main traffic in the commercial Recommender System in JD.com. To facilitate future research, we release the codes and datasets at https://github.com/guyulongcs/CIKM2020_DMT.

show abstract

“…Reference [22] introduced both character-level and work-level approaches to predict the CTR of a query-ad pair using deep convolutional neural networks where the textual content appearing in a query-ad pair and the page position on a search engine result page are only given as input. Ouyang et al examined various types of relationships including user-to-ad, ad-to-ad, and feature-to-CTR in [23], while, in [24], they tried to improve the CTR prediction of a target ad considering auxiliary ads from both the spatial domain and temporal domain. As in Web search, these studies are dedicated solutions tailored to the characteristics of sponsored search or contextual advertising and thus additional methods to relax the assumptions should be devised to utilize them for general Web applications.…”

Section: Web Navigation Prediction For Modern Web Applicationsmentioning

confidence: 99%

“…Second, navigation prediction for general Web applications has a low accuracy with conventional machine learning models. As described in Section II, researchers have studied user interaction prediction for Web applications focusing on two typical application types: Web search [9]- [16] and online advertising [5], [17], [18]- [24]. Although they provided in-depth insights on these traditional Web applications, today's complex Web environment with diverse applications, including augmented or virtual reality, finance, healthcare, social network, video streaming, and Web of Things, has not been studied well.…”

Section: Introductionmentioning

confidence: 99%

WebProfiler: User Interaction Prediction Framework for Web Applications

Joo

Lee

2019

IEEE Access

View full text Add to dashboard Cite

User interaction prediction for Web applications is crucial to improve browsing experience. With the prediction, for instance, a browser can prefetch the content to be accessed next reducing long wait times in advance. However, predicting the user interaction is challenging in practice. Collecting detailed interaction data is difficult due to the constraints on target application and platform. Moreover, Web navigation prediction mechanisms for general applications have a low accuracy with conventional machine learning models. To this end, in this paper, we propose a Web interaction profiling framework, WebProfiler, which collects user interaction data in a generic way and accurately predicts the next navigation. Both navigation and click events are collected by using JavaScript event handlers and clicked objects are identified reliably through a document object model based approach. Furthermore, we adopt gated recurrent unit (GRU), a representative deep learning technique suitable for coping with time series Web interaction data, and present two advanced techniques for training the GRU-based model: uniform resource locator (URL) grouping to handle the variant URLs of a Web page and Web embedding to represent both events in a unified vector space. The experimental results based on the real user interaction data showed that click events within an application improved the overall prediction performance by 13.7% on average, which were overlooked by most of the previous research. In addition, WebProfiler achieved an average F-measure of 0.798 for top three candidates where URL grouping and Web embedding contributed to 52.4% of the performance improvement. INDEX TERMS Deep learning, gated recurrent unit (GRU), navigation prediction, user interaction, web applications.

show abstract

Deep Spatio-Temporal Neural Networks for Click-Through Rate Prediction

Cited by 50 publications

References 26 publications

Modeling Users' Contextualized Page-wise Feedback for Click-Through Rate Prediction in E-commerce Search

Modeling Users' Contextualized Page-wise Feedback for Click-Through Rate Prediction in E-commerce Search

Deep Multifaceted Transformers for Multi-objective Ranking in Large-Scale E-commerce Recommender Systems

WebProfiler: User Interaction Prediction Framework for Web Applications

Contact Info

Product

Resources

About