ABSTRACT Many manufacturers have shifted from standalone products to Product‐Service Systems (PSS) to leverage flexible product‐service combinations and increase profit margins. However, the vast range of existing PSS design schemes makes it challenging to pinpoint personalized matches for diverse users. To address this issue, this study refines collaborative filtering (CF)‐based recommendation algorithms from a network perspective. Traditional CF relies solely on single‐layer user‐scheme relationships, reducing personalization. In contrast, we introduce a heterogeneous information network (HIN) to incorporate user‐user and scheme‐scheme similarity networks alongside user‐scheme relations. We then integrate HeteSim values from multiple search paths to more accurately compute user‐scheme preference relationships. In addition, four types of the core direct and indirect preference propagation paths between users and schemes are defined, and a parameter is introduced to penalize long paths, achieving a balance between computational efficiency and information loss. Considering the multi‐attribute nature of PSS design schemes, we combine subjective and objective weighting methods to adjust users' overall ratings, thereby capturing the multi‐dimensional attribute features of the schemes and improving the accuracy of recommendations. To validate its effectiveness, we applied our algorithm to new energy commercial van PSS (NEV‐PSS) recommendations, demonstrating superior personalized performance. The algorithm reveals implicit relationships between users and schemes, aiding suppliers in targeted marketing and informed decision‐making. It also helps predict user responses and market trends for new products, supporting technology development and regulatory policy recommendations.
Geng et al. (Fri,) studied this question.