Intent-aware knowledge graph-based model for electrical power material recommendation.

Area of Science:

• Electrical Power Material Management
• Recommender Systems
• Artificial Intelligence

Background:

• Accurate electrical material recommendations are crucial for effective management.
• Graph Neural Networks (GNNs) enhance recommendations by integrating node information and structure.
• Current GNN-based systems lack explicit user intent modeling, limiting performance.

Purpose of the Study:

• To propose an intent-aware knowledge graph-based model (IKG-EMR) for electrical material recommendation.
• To address the limitation of current GNNs in capturing user intent.
• To improve the accuracy and effectiveness of electrical material recommendations.

Main Methods:

• Developed the Intent-Aware Knowledge Graph-based Electrical Material Recommendation (IKG-EMR) model.
• Utilized a graph neural network (GNN) on a tripartite graph (User-Item-Topic) for intent and item embeddings.
• Employed a multi-head attention network (Transformer) to extract user preferences from behavior sequences.
• Integrated user preference and intent features using an adaptive fusion with attention network.

Main Results:

• The proposed IKG-EMR model effectively models user preferences and intent.
• Experimental results on real-life electric power materials demonstrate superior performance compared to state-of-the-art methods.
• IKG-EMR achieves enhanced recommendation accuracy by incorporating user intent.

Conclusions:

• The IKG-EMR model offers a significant advancement in electrical material recommendation systems.
• Explicitly modeling user intent is key to improving recommendation quality.
• The approach provides a robust framework for personalized recommendations in electrical power material management.