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

In the field of electrical power material management, it is paramount that users receive accurate recommendations regarding the electrical power materials they require. Recently, a growing number of studies have been dedicated to graph neural network (GNN)-based recommendation systems due to their ability to seamlessly combine node information with topological structure, enhancing the effectiveness of recommendations. However, a notable drawback of current GNN-based recommendation is their inability to explicitly capture users' intent in recommendations, which limits the performance. In fact, users' intent is crucial in determining their actions. One example is when users first form an intent to buy a particular set of items and then choose a specific item from the set based on their preferences. To fill this gap, this article proposes an intent-aware knowledge graph-based model for electrical material recommendation, named IKG-EMR. IKG-EMR models user preferences and intent by leveraging knowledge graph and user behavior sequences, respectively. Specifically, a graph neural network is adopted to generate user intent embedding and item embedding from the tripartite graph of "User-Item-Topic", and a multi-head attention network (Transformer) is used for extracting preference from user behavior sequences. Finally, an adaptive fusion with attention network is devised to generate comprehensive user representation by integrating user preference and intent features. Extensive experiments conducted on the real-life electric power materials show that our proposed model outperforms state-of-the-art methods.