Click Through Rate
Click-through rate (CTR) prediction aims to accurately estimate the probability of a user clicking on a given item, crucial for optimizing online advertising and recommender systems. Current research focuses on improving CTR prediction accuracy through advanced model architectures like Deep Interest Networks (DIN), transformers, and large language models (LLMs), often incorporating techniques such as feature selection, knowledge distillation, and bias mitigation to address challenges like data sparsity and position bias. These advancements significantly impact online businesses by enhancing ad targeting, personalization, and ultimately, revenue generation. Furthermore, ongoing work explores efficient training methods and addresses the complexities of handling massive datasets and concept drift in real-world applications.
Papers
On the Factory Floor: ML Engineering for Industrial-Scale Ads Recommendation Models
Rohan Anil, Sandra Gadanho, Da Huang, Nijith Jacob, Zhuoshu Li, Dong Lin, Todd Phillips, Cristina Pop, Kevin Regan, Gil I. Shamir, Rakesh Shivanna, Qiqi Yan
FiBiNet++: Reducing Model Size by Low Rank Feature Interaction Layer for CTR Prediction
Pengtao Zhang, Zheng Zheng, Junlin Zhang
Rethinking Position Bias Modeling with Knowledge Distillation for CTR Prediction
Congcong Liu, Yuejiang Li, Jian Zhu, Xiwei Zhao, Changping Peng, Zhangang Lin, Jingping Shao
On the Adaptation to Concept Drift for CTR Prediction
Congcong Liu, Yuejiang Li, Fei Teng, Xiwei Zhao, Changping Peng, Zhangang Lin, Jinghe Hu, Jingping Shao