Abstract
It is the manufacturer's responsibility to prevent microbial contamination during the cosmetic product life cycle; therefore, it is imperative that cosmetic formulations have adequate preservative efficacy. The preservative efficacy of a formulation is evaluated using preservative efficacy testing (PET), which usually requires a test period of about one month and failing PETs may lead to product launch delays. In order to detect formulations for which preservative measures had failed in the early stages of product development, we conducted a predictive analysis of preservative efficacy using artificial intelligence (AI). The training data for AI included formulation ingredients, such as the concentrations of ingredients of cosmetic leave-on care products like toners, emulsions, and creams, quasi-drugs, and pharmaceuticals as explanatory variables, and PET test results provided with Staphylococcus aureus as the objective variable. Among the formulation ingredients, the aqueous phase concentrations of preservatives, alkanediols with three or more carbon atoms, and ethanol, which act as preservative boosters, are theoretically calculated and they are used as explanatory variables. With the machine learning software “Prediction One”, a binary classification model was created to predict the efficacy of preservative approaches against S. aureus, and the prediction accuracy was verified by cross-validation. We were able to reach the possibility of predicting failures of preservative formulations against S. aureus with AI.
