Relationship Discovery and Hierarchical Embedding for Web Service Quality Prediction

Area of Science:

• Cloud Computing
• Internet of Things
• Artificial Intelligence
• Machine Learning

Background:

• Web Services Quality Prediction is crucial in Cloud Computing and IoT.
• Graph Convolutional Networks (GCNs) enhance prediction by aggregating local graph information.
• Existing GCN models struggle to fully utilize heterogeneous user-service bipartite graphs with multiple relationships.

Purpose of the Study:

• To propose a novel relationship discovery and hierarchical embedding method based on GCNs (RDHE).
• To address the limitations of previous GCNs in fully mining heterogeneous graph relationships.
• To develop a faster similarity calculation process for downstream prediction tasks.

Main Methods:

• Designed a dual mechanism for user and service representation.
• Developed a new community discovery method to leverage graph relationships.
• Implemented a fast similarity calculation process for efficient prediction.

Main Results:

• The RDHE method demonstrated significant improvements in web service quality prediction accuracy.
• Experiments on real datasets validated the effectiveness of mining diverse relationships in heterogeneous graphs.
• The proposed method efficiently utilizes graph structures for enhanced prediction performance.

Conclusions:

• The RDHE method offers a more comprehensive approach to Web Services Quality Prediction by fully exploiting heterogeneous graph relationships.
• This work advances GCN applications in Cloud Computing and IoT by improving prediction accuracy and efficiency.
• The developed techniques provide a robust framework for complex graph-based prediction tasks.