In this study, a CatBoost regression model was constructed to clarify the relationship between geometric features extracted from tire tread images and sensory evaluations of “quietness” and “grip performance” on DRY and WET surfaces. SHapley Additive exPlanations (SHAP) was then applied to visualize the contribution of each feature, enabling the identification of geometric features that are important for sensory evaluation. Furthermore, by comparing the sensory evaluations of general users and tire manufacturer employees, we identified the geometric features that caused differences in their evaluations.
KODA et al. (Wed,) studied this question.