Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

GACE: Learning Graph-Based Cross-Page Ads Embedding For Click-Through Rate Prediction

Haowen Wang, Yuliang Du, Congyun Jin, Yujiao Li, Yingbo Wang, Tao Sun, Piqi Qin, Cong Fan

TL;DR

GACE addresses CTR prediction for cross-page ads, focusing on cold-start and cross-page data challenges. It builds a weighted undirected graph from semantic, user interaction, and page knowledge, and learns ad embeddings through a variational graph auto-encoder with graph attention. The method demonstrates superior offline and online performance, notably improving cold-start CTR across multiple base models and pages on AliEC and Alipay datasets. This cross-page embedding approach offers practical gains for real-world ad recommendation systems and can be extended to additional metrics such as CVR and GMV.

Abstract

Predicting click-through rate (CTR) is the core task of many ads online recommendation systems, which helps improve user experience and increase platform revenue. In this type of recommendation system, we often encounter two main problems: the joint usage of multi-page historical advertising data and the cold start of new ads. In this paper, we proposed GACE, a graph-based cross-page ads embedding generation method. It can warm up and generate the representation embedding of cold-start and existing ads across various pages. Specifically, we carefully build linkages and a weighted undirected graph model considering semantic and page-type attributes to guide the direction of feature fusion and generation. We designed a variational auto-encoding task as pre-training module and generated embedding representations for new and old ads based on this task. The results evaluated in the public dataset AliEC from RecBole and the real-world industry dataset from Alipay show that our GACE method is significantly superior to the SOTA method. In the online A/B test, the click-through rate on three real-world pages from Alipay has increased by 3.6%, 2.13%, and 3.02%, respectively. Especially in the cold-start task, the CTR increased by 9.96%, 7.51%, and 8.97%, respectively.

GACE: Learning Graph-Based Cross-Page Ads Embedding For Click-Through Rate Prediction

TL;DR

GACE addresses CTR prediction for cross-page ads, focusing on cold-start and cross-page data challenges. It builds a weighted undirected graph from semantic, user interaction, and page knowledge, and learns ad embeddings through a variational graph auto-encoder with graph attention. The method demonstrates superior offline and online performance, notably improving cold-start CTR across multiple base models and pages on AliEC and Alipay datasets. This cross-page embedding approach offers practical gains for real-world ad recommendation systems and can be extended to additional metrics such as CVR and GMV.

Abstract

Predicting click-through rate (CTR) is the core task of many ads online recommendation systems, which helps improve user experience and increase platform revenue. In this type of recommendation system, we often encounter two main problems: the joint usage of multi-page historical advertising data and the cold start of new ads. In this paper, we proposed GACE, a graph-based cross-page ads embedding generation method. It can warm up and generate the representation embedding of cold-start and existing ads across various pages. Specifically, we carefully build linkages and a weighted undirected graph model considering semantic and page-type attributes to guide the direction of feature fusion and generation. We designed a variational auto-encoding task as pre-training module and generated embedding representations for new and old ads based on this task. The results evaluated in the public dataset AliEC from RecBole and the real-world industry dataset from Alipay show that our GACE method is significantly superior to the SOTA method. In the online A/B test, the click-through rate on three real-world pages from Alipay has increased by 3.6%, 2.13%, and 3.02%, respectively. Especially in the cold-start task, the CTR increased by 9.96%, 7.51%, and 8.97%, respectively.
Paper Structure (27 sections, 15 equations, 3 figures, 5 tables)

This paper contains 27 sections, 15 equations, 3 figures, 5 tables.

Table of Contents

  1. Introduction
  2. RELATED WORKS
  3. General CTR recommendation systems
  4. Item Embedding Generation
  5. PRELIMINARIES
  6. Item Knowledge Base
  7. Semantic Knowledge
  8. User Interaction Knowledge
  9. Page Knowledge
  10. Problem Formulation
  11. METHODOLOGY
  12. Overview
  13. Graph Creation
  14. Pre-training Embedding Learning
  15. Experiment
  16. ...and 12 more sections

Figures (3)

  • Figure 1: Item Knowledge Base: semantic knowledge, user interaction knowledge and page knowledge
  • Figure 2: The graph creation part for GACE using item knowledge base
  • Figure 3: The workflow for self-encode task based on variational graph auto-encoder.