Construction of a user clothing preference prediction model based on KCCF and BP neural networks

Purpose – To more accurately predict user preferences in the clothing industry, a user clothing preference prediction model combining the KCCF model and BP neural network model was proposed. Design/methodology/approach – Using men's suits as an example, the model first constructed a Kansei space using semantic differential and factor analysis methods, establishing 8 pairs of representative Kansei words. The shape, color, and texture dimensions of 100 men's suits were then quantified. Next, a coupling coordination model was used to establish the coordination relationship between these dimensions and complete their quantification. Finally, a 7 × 15 × 1 BP neural network model was constructed to predict user preferences and measure the model's accuracy. Findings – The results were compared with traditional one-dimensional and the proposed three-dimensional models. The 7-dimensional model performed exceptionally well on all prediction accuracy metrics, significantly improving prediction accuracy. Originality/value – The combined KCCF and BP neural network prediction model demonstrates strong operability, not only effectively supporting designers in predicting user preferences, but also providing guidance for scheme selection, design optimization, and market decision-making in clothing product development.