Research on cloud manufacturing service recommendation based on graph neural network.

Area of Science:

• Computer Science
• Manufacturing Engineering
• Artificial Intelligence

Background:

• Cloud manufacturing (CMfg) platforms face information overload due to a growing number of service resources.
• Traditional recommendation methods struggle to efficiently identify suitable CMfg resources.

Purpose of the Study:

• To propose a novel graph neural network (GNN)-based recommendation method for CMfg service resources.
• To address the information overloading problem for enterprises seeking CMfg services.

Main Methods:

• Constructing a resource graph dataset using similarity metrics (Cosine, Pearson) on CMfg service features.
• Applying GNNs for representation learning to obtain vector embeddings of resources.
• Predicting potential resource links via dot product calculations on node embeddings for recommendations.

Main Results:

• The proposed GNN method demonstrated superior link prediction accuracy compared to existing algorithms.
• Predictive performance significantly improved with higher network density during training.

Conclusions:

• The GNN-based approach effectively recommends suitable CMfg service resources.
• This method offers a viable solution for navigating complex CMfg service platforms.