An Intelligent Framework for Multi-Scenario Fabric Selection in Cycling Apparel: A User-Expert Dual-Driven Approach

Abstract

Traditional evaluation of cycling trouser fabrics employs fixed-weight methods, failing to address the diverse performance demands of dynamic cycling scenarios. To overcome this limitation, this study developed and validated an intelligent scoring system using a “user-expert” dual-driven dynamic weight optimisation strategy. A dynamic weighting scheme was first constructed by integrating the Analytic Hierarchy Process (AHP) with user survey data. To mitigate data scarcity, a Generative Adversarial Network (GAN) was utilised to augment the dataset. Subsequently, a random forest regression model, integrated with a fuzzy logic engine, was employed to dynamically adjust performance weights (abrasion resistance, moisture permeability, air permeability) based on scenario inputs. The model achieved a coefficient of determination ($R^2$) of 0.924 on the training set and an average $R^2$ of 0.893 on the test set. Further validation demonstrated a prediction error of $\le 5.2\%$ for new fabrics and maintained $R^2$ values above 0.85 under hyperparameter sensitivity analysis, indicating strong generalisation and robustness. This research provides a validated, data-driven framework for fabric selection in customised clothing and introduces a novel paradigm for the performance-based design of advanced technical textiles.