Prediction of treatment effect perception in cosmetics using machine learning.

Area of Science:

• Cosmetic Science
• Dermatology
• Image Analysis

Background:

• Perception of cosmetic treatment effect (TE) is complex, influenced by multiple factors.
• Simultaneous consideration of various parameters is crucial for accurate TE evaluation.
• Existing methods may not fully capture the multifaceted nature of TE perception.

Purpose of the Study:

• To develop and validate a global approach for predicting TE perception in cosmetic products.
• To utilize a Random Forest (RF) classifier for modeling TE perception.
• To enhance the substantiation of consumer-centered claims in clinical trials.

Main Methods:

• Employed data from three randomized, double-blind clinical studies (n=50).
• Assessed nine primary endpoints related to skin pores using image analysis algorithms.
• Utilized a Random Forest (RF) classifier to predict TE perception based on endpoints and expert evaluations.
• Applied the Bradley-Terry model for intra-study product ranking based on judge preferences.

Main Results:

• The RF model demonstrated good accuracy in predicting TE perception.
• The RF approach effectively handled multiplicity, nonlinearity, and interactions among criteria.
• The method simplifies decision-making by focusing on a discrete parameter.

Conclusions:

• Random Forest (RF) offers a robust and accurate method for predicting cosmetic treatment effect (TE) perception.
• This approach simplifies interpretability and supports consumer-centered claim substantiation.
• The developed model provides a valuable tool for evaluating cosmetic product efficacy.