The breakthrough and imagination of advertising depth prediction technology in the scenario of Meituan to the store
In the post-deep learning era, technology iteration has fully entered the deep water zone, and the optimization of advertising prediction models that mainly increase model complexity is no longer effective. The Meituan to-store advertising quality estimation team closely combined with the characteristics of the business and took advantage of the flexible and changeable structure of the in-depth model to achieve further breakthroughs. This article first introduces the two major challenges of Meituan’s LBS space distance constraint and long-term periodicity, and then introduces the four-dimensional response solutions of context, users, advertising, and training methods. The specific four technological breakthroughs are as follows: a. Based on ranking Combined context bias perception prediction; b. Super-long sequence modeling based on time-space dependence; c. Advertising candidate dynamics; d. Disaster forgetting and continuous learning, while driving significant improvements in online indicators, it is organized as a paper and published in SIGIR , CIKM and other top international conferences. Finally, based on a new round of understanding, new trends in forecasting technologies such as dynamic inference levels and differentiated evaluation indicators are proposed.
-  Friedman J H . Greedy Function Approximation: A Gradient Boosting Machine[J]. Annals of Statistics, 2001, 29(5):1189-1232.
-  Rendle S. Factorization machines[C]//2010 IEEE International conference on data mining. IEEE, 2010: 995-1000.
-  HT Cheng, et al. Wide & Deep Learning for Recommender Systems, 2016
-  Zhou, Guorui, et al. “Deep interest network for click-through rate prediction.” Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining. 2018
-  Modeling Task Relationships in Multi-task Learning with Multi-gate Mixture-of-Experts. ACM, 2018.
-  Wen, Ling, Chua. A closer look at strategies for memorization.[J]. Clavier Companion, 2014, 6(6):50-52.
-  Huang J, Hu K, Tang Q, et al. Deep Position-wise Interaction Network for CTR Prediction[J]. arXiv preprint arXiv:2106.05482, 2021.
-  Search-based User Interest Modeling with Lifelong Sequential Behavior Data for Click-Through Rate Prediction
-  Qi, Yi, et al. "Trilateral Spatiotemporal Attention Network for User Behavior Modeling in Location-based Search", CIKM 2021.
-  Overcoming catastrophic forgetting in neural networks[J]. Proceedings of the National Academy of Sciences of the United States of America, 2017.
-  M. Zinkevich. Online convex programming and generalized infinitesimal gradient ascent. In ICML, 2003.
-  Optimized Cost per Click in Taobao Display Advertising[C]// the 23rd ACM SIGKDD International Conference. ACM, 2017.
-  https://tech.meituan.com/2020/08/20/kdd-cup-debiasing-practice.html
The advertising algorithm team of the Meituan to-store advertising platform is based on advertising scenarios, exploring the development of cutting-edge technology in deep learning, reinforcement learning, artificial intelligence, big data, knowledge graphs, NLP and computer vision, and exploring the value of local life service e-commerce. The main work directions include:
- trigger strategy : user intention recognition, advertising merchant data understanding, Query rewriting, deep matching, and correlation modeling.
- Quality estimation : Modeling the quality of advertising. Estimated click-through rate, conversion rate, customer unit price, and transaction volume.
- mechanism design : advertising ranking mechanism, bidding mechanism, bid suggestion, traffic estimation, budget allocation.
- creative optimization : intelligent creative design. Optimize the display creativity of advertising pictures, text, group orders, and preferential information.
- Have more than three years of relevant work experience, and have application experience in at least one aspect of CTR/CVR estimation, NLP, image understanding, and mechanism design.
- Familiar with commonly used machine learning, deep learning, and reinforcement learning models.
- Excellent logical thinking ability, passion for solving challenging problems, sensitive to data, and good at analyzing/solving problems.
- Master degree or above in computer and mathematics related majors.
following conditions to give priority to :
- Have relevant business experience in advertising/search/recommendation.
- Have experience in large-scale machine learning.
Interested students can submit their resumes to: email@example.com (please indicate the subject of the email: Meituan Guangping Algorithm Team).
Read more technical articles from the
| In the public account menu bar dialog box, reply to keywords such as [2020 goods], [2019 goods], [2018 goods], and [2017 goods]. You can view the collection of technical articles from the Meituan technical team over the years.
| This article is produced by the Meituan technical team, and the copyright belongs to Meituan. Welcome to reprint or use the content of this article for non-commercial purposes such as sharing and communication, please indicate "the content is reproduced from the Meituan technical team". This article may not be reproduced or used commercially without permission. For any commercial activity, please send an email to firstname.lastname@example.org to apply for authorization.