A session-based recommendation method for user-centric new product design
Jun Wu, Xinyi Lan, Xianyi Chen · Journal of Engineering Design · 2025
In the current competitive market environment, user-centric product design is gaining prominence. As a significant task, user-centric product design needs to reuse existing product data and information, including user feedback, product services, and user preferences, to assist product design decision-making. However, new product design is challenged by the absence of historical user interaction data. In this paper, a Session-based Decision Recommendation (SDR) method is proposed for the design of a user-centric new product. This method employs a self-supervised model to extract user preferences and uses a contrastive alignment module to capture the complex nonlinear relationships between user preferences and new product designs. Specifically, it enhances session graph data through edge reweighting rather than random deletion, effectively alleviating data sparsity while preserving the critical graph structure. Inspired by graph contrastive alignment, our method constructs a novel contrastive alignment module, which can learn complex interactions between transformed product attribute representations and actual product representations, thereby improving alignment between new product design representations and user preferences. Extensive experiments on two real-world datasets demonstrate the effectiveness of our method. The findings of the current study highlight the determinant role of session-based decision recommendation in developing new product design.