Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Deep Group Interest Modeling of Full Lifelong User Behaviors for CTR Prediction

Qi Liu, Xuyang Hou, Haoran Jin, Xiaolong Chen, Jin Chen, Defu Lian, Zhe Wang, Jia Cheng, Jun Lei

TL;DR

This research introduces the Deep Group Interest Network (DGIN), an end-to-end method to model the user's entire behavior history, providing a nuanced user understanding and validates DGIN's efficacy and efficiency.

Abstract

Extracting users' interests from their lifelong behavior sequence is crucial for predicting Click-Through Rate (CTR). Most current methods employ a two-stage process for efficiency: they first select historical behaviors related to the candidate item and then deduce the user's interest from this narrowed-down behavior sub-sequence. This two-stage paradigm, though effective, leads to information loss. Solely using users' lifelong click behaviors doesn't provide a complete picture of their interests, leading to suboptimal performance. In our research, we introduce the Deep Group Interest Network (DGIN), an end-to-end method to model the user's entire behavior history. This includes all post-registration actions, such as clicks, cart additions, purchases, and more, providing a nuanced user understanding. We start by grouping the full range of behaviors using a relevant key (like item_id) to enhance efficiency. This process reduces the behavior length significantly, from O(10^4) to O(10^2). To mitigate the potential loss of information due to grouping, we incorporate two categories of group attributes. Within each group, we calculate statistical information on various heterogeneous behaviors (like behavior counts) and employ self-attention mechanisms to highlight unique behavior characteristics (like behavior type). Based on this reorganized behavior data, the user's interests are derived using the Transformer technique. Additionally, we identify a subset of behaviors that share the same item_id with the candidate item from the lifelong behavior sequence. The insights from this subset reveal the user's decision-making process related to the candidate item, improving prediction accuracy. Our comprehensive evaluation, both on industrial and public datasets, validates DGIN's efficacy and efficiency.

Deep Group Interest Modeling of Full Lifelong User Behaviors for CTR Prediction

TL;DR

This research introduces the Deep Group Interest Network (DGIN), an end-to-end method to model the user's entire behavior history, providing a nuanced user understanding and validates DGIN's efficacy and efficiency.

Abstract

Extracting users' interests from their lifelong behavior sequence is crucial for predicting Click-Through Rate (CTR). Most current methods employ a two-stage process for efficiency: they first select historical behaviors related to the candidate item and then deduce the user's interest from this narrowed-down behavior sub-sequence. This two-stage paradigm, though effective, leads to information loss. Solely using users' lifelong click behaviors doesn't provide a complete picture of their interests, leading to suboptimal performance. In our research, we introduce the Deep Group Interest Network (DGIN), an end-to-end method to model the user's entire behavior history. This includes all post-registration actions, such as clicks, cart additions, purchases, and more, providing a nuanced user understanding. We start by grouping the full range of behaviors using a relevant key (like item_id) to enhance efficiency. This process reduces the behavior length significantly, from O(10^4) to O(10^2). To mitigate the potential loss of information due to grouping, we incorporate two categories of group attributes. Within each group, we calculate statistical information on various heterogeneous behaviors (like behavior counts) and employ self-attention mechanisms to highlight unique behavior characteristics (like behavior type). Based on this reorganized behavior data, the user's interests are derived using the Transformer technique. Additionally, we identify a subset of behaviors that share the same item_id with the candidate item from the lifelong behavior sequence. The insights from this subset reveal the user's decision-making process related to the candidate item, improving prediction accuracy. Our comprehensive evaluation, both on industrial and public datasets, validates DGIN's efficacy and efficiency.
Paper Structure (25 sections, 8 equations, 5 figures, 5 tables)

This paper contains 25 sections, 8 equations, 5 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Click-Through Rate Prediction
  4. Long User Behavior Sequence Modeling
  5. Methods
  6. Preliminaries
  7. Embedding Layer
  8. Group Module
  9. Statistical Attributes
  10. Aggregated Attribute
  11. Attention On Interest Set
  12. Target Module
  13. Experiment Setup
  14. Datasets
  15. Baselines
  16. ...and 10 more sections

Figures (5)

  • Figure 1: The distribution of behaviors' length affected by grouping on item_id. We rank samples on behaviors' length in ascending order. The left half shows the corresponding behavior's length at different quantiles. The right half displays the distribution of the behavior quantity of each group.
  • Figure 2: The overall framework of Deep Group Interest Network (DGIN). DGIN consists of Group Module (GM) and Target Module (TM), which capture long-term and decision interest from lifelong user multi-behavior sequence.
  • Figure 3: The detail of GM's architecture. We take item_id as the interest key for illustration.
  • Figure 4: The detail of TM's architecture.
  • Figure 5: The whole system architecture for deploying DGIN.