Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Coarse-to-fine Dynamic Uplift Modeling for Real-time Video Recommendation

Chang Meng, Chenhao Zhai, Xueliang Wang, Shuchang Liu, Xiaoqiang Feng, Lantao Hu, Xiu Li, Han Li, Kun Gai

TL;DR

This work designs adjusting the distribution of videos with varying durations as the treatment and proposes Coarse-to-fine Dynamic Uplift Modeling (CDUM) for real-time video recommendation, which is eventually fully deployed on the Kuaishou platform.

Abstract

With the rise of short video platforms, video recommendation technology faces more complex challenges. Currently, there are multiple non-personalized modules in the video recommendation pipeline that urgently need personalized modeling techniques for improvement. Inspired by the success of uplift modeling in online marketing, we attempt to implement uplift modeling in the video recommendation scenario. However, we face two main challenges: 1) Design and utilization of treatments, and 2) Capture of user real-time interest. To address them, we design adjusting the distribution of videos with varying durations as the treatment and propose Coarse-to-fine Dynamic Uplift Modeling (CDUM) for real-time video recommendation. CDUM consists of two modules, CPM and FIC. The former module fully utilizes the offline features of users to model their long-term preferences, while the latter module leverages online real-time contextual features and request-level candidates to model users' real-time interests. These two modules work together to dynamically identify and targeting specific user groups and applying treatments effectively. Further, we conduct comprehensive experiments on the offline public and industrial datasets and online A/B test, demonstrating the superiority and effectiveness of our proposed CDUM. Our proposed CDUM is eventually fully deployed on the Kuaishou platform, serving hundreds of millions of users every day. The source code will be provided after the paper is accepted.

Coarse-to-fine Dynamic Uplift Modeling for Real-time Video Recommendation

TL;DR

This work designs adjusting the distribution of videos with varying durations as the treatment and proposes Coarse-to-fine Dynamic Uplift Modeling (CDUM) for real-time video recommendation, which is eventually fully deployed on the Kuaishou platform.

Abstract

With the rise of short video platforms, video recommendation technology faces more complex challenges. Currently, there are multiple non-personalized modules in the video recommendation pipeline that urgently need personalized modeling techniques for improvement. Inspired by the success of uplift modeling in online marketing, we attempt to implement uplift modeling in the video recommendation scenario. However, we face two main challenges: 1) Design and utilization of treatments, and 2) Capture of user real-time interest. To address them, we design adjusting the distribution of videos with varying durations as the treatment and propose Coarse-to-fine Dynamic Uplift Modeling (CDUM) for real-time video recommendation. CDUM consists of two modules, CPM and FIC. The former module fully utilizes the offline features of users to model their long-term preferences, while the latter module leverages online real-time contextual features and request-level candidates to model users' real-time interests. These two modules work together to dynamically identify and targeting specific user groups and applying treatments effectively. Further, we conduct comprehensive experiments on the offline public and industrial datasets and online A/B test, demonstrating the superiority and effectiveness of our proposed CDUM. Our proposed CDUM is eventually fully deployed on the Kuaishou platform, serving hundreds of millions of users every day. The source code will be provided after the paper is accepted.

Paper Structure

This paper contains 29 sections, 9 equations, 4 figures, 5 tables.

Table of Contents

  1. INTRODUCTION
  2. RELATED WORK
  3. PRELIMINARIES
  4. Offline Problem Definition
  5. Online Problem Definition
  6. METHOD
  7. Coarse-grained Preference Modeling
  8. Feature Encoder Module.
  9. The Multi-treatment Learning Module
  10. Training Objective and Prediction.
  11. Fine-grained Interest Capture
  12. Module Structue.
  13. Training Objective and Prediction
  14. Decision Making.
  15. OFFLINE EVALUATIONS
  16. ...and 14 more sections

Figures (4)

  • Figure 1: A example of the difference between online marketing and video recommendation scenarios.
  • Figure 2: Illustration of the proposed CDUM framework. The light purple block on the left is the offline training part (CPM), and the light green block on the right is the online inference part (FIC).
  • Figure 3: Illustration of the online pipeline.
  • Figure 4: Online A/B test experimental results of Enter LT and Slide LT.