.png){kind=icon}
.png){kind=icon}
Return
Explainable knowledge recommendation for product innovation concept design based on knowledge graph and multi-task learning
.png){kind=icon}
PRE
AI
DOI：10.1016/j.engappai.2025.112231.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
In the process of product innovation concept design, by enhancing the transparency of the knowledge recommendation process and providing designers with explainable recommendation results, decision-making time can be reduced, and design efficiency can be improved. Existing explainable knowledge recommendation methods can generate textual explanation information, but they often overlook the dynamic nature of the design process, which negatively affects the accuracy of the recommendations. To address this issue, the current study proposes a knowledge recommendation method for product innovation concept design based on knowledge graph (KG) and multi-task learning. Specifically, a TransD-based model is first constructed to perform the knowledge graph embedding (KGE) task. Then, KGE and Gated Recurrent Unit (GRU) are used to obtain dynamic knowledge demands across different temporal scales from the historical interactions of designers to enhance recommendation accuracy. A multi-task learning framework is subsequently introduced to enable feature sharing between the two tasks, improving the generalization and effectiveness of the model. Finally, an explanation strategy is designed by combining embedded knowledge and paths to provide optimal explainability information. Experimental results demonstrate that, compared with other state-of-the-art knowledge recommendation algorithms, the proposed method not only offers explainable recommendation results but also achieves higher accuracy, making it more suitable for real-world recommendation scenarios.
Journal
IF:
8
Papers:
5.4K
Citations:
3.5W