Learning design preferences through design feature extraction and weighted ensemble

Abstract Design plays a critical role in consumer purchase decisions. As the need for understanding and predicting various preferences for each customer increases, along with the importance of mass customization, predicting individual design preferences has become a critical factor in product development. However, current methods for predicting design preferences have several limitations. Product design involves a vast amount of high-dimensional information, and individual design preferences represent complex, heterogeneous emotional responses unique to each person. To address these challenges, we propose an approach that utilizes a dimensionality reduction model to transform design samples into low-dimensional feature vectors, enabling us to extract the key representative features of each design. By leveraging the design preference tendencies of others within our preference prediction models, we can predict individual-level design preferences more accurately. Our proposed framework overcomes the limitations of traditional methods for determining design preferences, allowing us to accurately identify design features and predict individual preferences for specific products. This framework improves the effectiveness of product development and enables personalized product recommendations that cater to the unique needs of each consumer.