Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Latent Factor Modeling with Expert Network for Multi-Behavior Recommendation

Mingshi Yan, Zhiyong Cheng, Yahong Han, Meng Wang

Abstract

Traditional recommendation methods, which typically focus on modeling a single user behavior (e.g., purchase), often face severe data sparsity issues. Multi-behavior recommendation methods offer a promising solution by leveraging user data from diverse behaviors. However, most existing approaches entangle multiple behavioral factors, learning holistic but imprecise representations that fail to capture specific user intents. To address this issue, we propose a multi-behavior method by modeling latent factors with an expert network (MBLFE). In our approach, we design a gating expert network, where the expert network models all latent factors within the entire recommendation scenario, with each expert specializing in a specific latent factor. The gating network dynamically selects the optimal combination of experts for each user, enabling a more accurate representation of user preferences. To ensure independence among experts and factor consistency of a particular expert, we incorporate self-supervised learning during the training process. Furthermore, we enrich embeddings with multi-behavior data to provide the expert network with more comprehensive collaborative information for factor extraction. Extensive experiments on three real-world datasets demonstrate that our method significantly outperforms state-of-the-art baselines, validating its effectiveness.

Latent Factor Modeling with Expert Network for Multi-Behavior Recommendation

Abstract

Traditional recommendation methods, which typically focus on modeling a single user behavior (e.g., purchase), often face severe data sparsity issues. Multi-behavior recommendation methods offer a promising solution by leveraging user data from diverse behaviors. However, most existing approaches entangle multiple behavioral factors, learning holistic but imprecise representations that fail to capture specific user intents. To address this issue, we propose a multi-behavior method by modeling latent factors with an expert network (MBLFE). In our approach, we design a gating expert network, where the expert network models all latent factors within the entire recommendation scenario, with each expert specializing in a specific latent factor. The gating network dynamically selects the optimal combination of experts for each user, enabling a more accurate representation of user preferences. To ensure independence among experts and factor consistency of a particular expert, we incorporate self-supervised learning during the training process. Furthermore, we enrich embeddings with multi-behavior data to provide the expert network with more comprehensive collaborative information for factor extraction. Extensive experiments on three real-world datasets demonstrate that our method significantly outperforms state-of-the-art baselines, validating its effectiveness.
Paper Structure (30 sections, 22 equations, 5 figures, 4 tables)

This paper contains 30 sections, 22 equations, 5 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Multi-behavior Recommendation
  4. Disentangled Representation Learning
  5. Mixture of Experts in Recommendation
  6. Methodology
  7. Preliminaries
  8. MBLFE Method
  9. Embedding Enhancement Network
  10. Gating Expert Network
  11. Target Behavior Recommendation
  12. Model Training
  13. Model Analysis
  14. Time Complexity Analysis
  15. Discussion
  16. ...and 15 more sections

Figures (5)

  • Figure 1: An illustrating of the latent factors.
  • Figure 2: Overview of MBLFE. (The left side shows the overall architecture of MBLFE, while the right side illustrates the internal details of the gating expert network. The initialized embedding is first enhanced by the embedding enhancement network. Then, the gating expert network separately extracts latent factors for both users and items. Finally, the user's latent factors are projected into the target behavior space to generate recommendations. Users and items share a common gating expert network.)
  • Figure 3: Statistics on the number of experts selected by users (the $x$-$axis$ and $y$-$axis$ represent the number of experts and the number of users, respectively, and "# Experts=18" means that the total number of experts is 18).
  • Figure 4: Visualization of latent factors distribution extracted by different experts (each color corresponds to the factor extracted by a specific expert, and "# Experts=18" means that the total number of experts is 18).
  • Figure 5: Impact of the total number of experts on model performance.