Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

QoS-Aware Graph Contrastive Learning for Web Service Recommendation

Jeongwhan Choi, Duksan Ryu

TL;DR

This work tackles the challenges of data sparsity and cold-start in QoS-based web service recommendation by proposing QAGCL, a QoS-aware graph contrastive learning framework. QAGCL constructs contextually augmented graphs that incorporate geolocation information and randomness via Haversine-distance masking and edge dropout, and learns user/service embeddings through graph convolutions. A contrastive learning objective (InfoNCE) aligns representations across the augmented views, while a Bayesian Personalized Ranking objective guides final predictions, yielding improved top-K recall and NDCG on the WSDream dataset across warm-start and cold-start scenarios. The approach demonstrates strong robustness to sparse data and uncertainty, with ablations showing the value of combining distance-aware and stochastic graph augmentations. Overall, QAGCL advances QoS-aware service recommendation by leveraging contextual graph augmentation and contrastive learning for more accurate, scalable recommendations in real-world settings.

Abstract

With the rapid growth of cloud services driven by advancements in web service technology, selecting a high-quality service from a wide range of options has become a complex task. This study aims to address the challenges of data sparsity and the cold-start problem in web service recommendation using Quality of Service (QoS). We propose a novel approach called QoS-aware graph contrastive learning (QAGCL) for web service recommendation. Our model harnesses the power of graph contrastive learning to handle cold-start problems and improve recommendation accuracy effectively. By constructing contextually augmented graphs with geolocation information and randomness, our model provides diverse views. Through the use of graph convolutional networks and graph contrastive learning techniques, we learn user and service embeddings from these augmented graphs. The learned embeddings are then utilized to seamlessly integrate QoS considerations into the recommendation process. Experimental results demonstrate the superiority of our QAGCL model over several existing models, highlighting its effectiveness in addressing data sparsity and the cold-start problem in QoS-aware service recommendations. Our research contributes to the potential for more accurate recommendations in real-world scenarios, even with limited user-service interaction data.

QoS-Aware Graph Contrastive Learning for Web Service Recommendation

TL;DR

This work tackles the challenges of data sparsity and cold-start in QoS-based web service recommendation by proposing QAGCL, a QoS-aware graph contrastive learning framework. QAGCL constructs contextually augmented graphs that incorporate geolocation information and randomness via Haversine-distance masking and edge dropout, and learns user/service embeddings through graph convolutions. A contrastive learning objective (InfoNCE) aligns representations across the augmented views, while a Bayesian Personalized Ranking objective guides final predictions, yielding improved top-K recall and NDCG on the WSDream dataset across warm-start and cold-start scenarios. The approach demonstrates strong robustness to sparse data and uncertainty, with ablations showing the value of combining distance-aware and stochastic graph augmentations. Overall, QAGCL advances QoS-aware service recommendation by leveraging contextual graph augmentation and contrastive learning for more accurate, scalable recommendations in real-world settings.

Abstract

With the rapid growth of cloud services driven by advancements in web service technology, selecting a high-quality service from a wide range of options has become a complex task. This study aims to address the challenges of data sparsity and the cold-start problem in web service recommendation using Quality of Service (QoS). We propose a novel approach called QoS-aware graph contrastive learning (QAGCL) for web service recommendation. Our model harnesses the power of graph contrastive learning to handle cold-start problems and improve recommendation accuracy effectively. By constructing contextually augmented graphs with geolocation information and randomness, our model provides diverse views. Through the use of graph convolutional networks and graph contrastive learning techniques, we learn user and service embeddings from these augmented graphs. The learned embeddings are then utilized to seamlessly integrate QoS considerations into the recommendation process. Experimental results demonstrate the superiority of our QAGCL model over several existing models, highlighting its effectiveness in addressing data sparsity and the cold-start problem in QoS-aware service recommendations. Our research contributes to the potential for more accurate recommendations in real-world scenarios, even with limited user-service interaction data.
Paper Structure (33 sections, 18 equations, 4 figures, 6 tables, 1 algorithm)

This paper contains 33 sections, 18 equations, 4 figures, 6 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Prelimaries & Related Work
  3. Web Quality of Service (QoS)
  4. Web Service Recommendation
  5. Graph-based Collaborative Filtering
  6. Contrastive Learning for Recommendation
  7. Motivation
  8. Why are GCNs suitable for CF?
  9. Why is context information such as geolocation important?
  10. Why is it appropriate to perform contrastive learning by creating augmented views in utilizing context information?
  11. Methododology
  12. Overall Architecture
  13. Graph Augmentation
  14. Haversine Distance (HD)
  15. Edge Dropout (ED)
  16. ...and 18 more sections

Figures (4)

  • Figure 1: Illustration of the predictions of QoS from two models. The black edge is the response time, which is the QoS. The blue dashed edge represents the predicted response time of $M_1$ model, and the red dashed edge represents the predicted response time of $M_2$ model.
  • Figure 2: A framework of our proposed QAGCL
  • Figure 3: Illustration of the Haversine distance and straight line distance between two points on a sphere.
  • Figure 4: The geolocations information of WSDream. The red (resp. blue) circles are the users' (resp. services') location. The black lines are interactions between users and services.