Predictive models for the sensory textures of cosmetic formulations based on rheological measurements
발표자
()
초록
내용
Sensory properties of cosmetics are important factors that affect consumer preferences. Conventional sensory analyses that based on panel tests have limitations in that they are time-consuming and costly. As an alternative to panel-based methods, we integrate rheological measurements and machine learning techniques to offer prediction models for several sensory texture (Adhesiveness, Softness, Stickiness, and Thickness). Predictive models were created using the random forest regressior and the rheology-panel test library of over 100 cosmetic formulas. As feature variables, metrics from linear, nonlinear, and extensional rheological measurements are employed. We chose the features from each sort of measurement that had the highest feature relevance to come up with the best feature set. Our results indicate that sensory properties of cosmetics can be effectively estimated from rheological measurements. Moreover, our prediction models explain which rheological characteristic is crucial in determining each sensory texture.