The purpose of this study is to predict tactile sensation for velvet cosmetic puffs. A device equipped with a rotational mechanism was developed to simultaneously measure the compression and friction properties of anisotropic velvet fabrics used in cosmetic powder puffs. This mechanism enables efficient multi-directional evaluation and extraction of mechanical features. A random forest regression model was applied to predict sensory values from these features. Sensory prediction was performed using average sensory evaluations. This approach was adopted because the inter-subject variation was relatively small within a seven-point rating scale, making the average value an appropriate prediction target. The prediction accuracy of the model was verified using untrained samples and remained high even when different subject groups were used for training and validation. These results demonstrate the effectiveness of the proposed method as a cost-efficient alternative to sensory testing and its potential for tactile evaluation and product development of cosmetic applicators.
Oda et al. (Thu,) studied this question.