z-logo
Premium
Online comment‐based prediction of cosmetic ingredient's sensory irritation using gradient boosting algorithm
Author(s) -
Jiang Biao,
Wang Huijuan,
Cheng Li,
Zi Yusha,
He Congfen,
Den YiAnn
Publication year - 2020
Publication title -
journal of cosmetic dermatology
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.626
H-Index - 44
eISSN - 1473-2165
pISSN - 1473-2130
DOI - 10.1111/jocd.13201
Subject(s) - boosting (machine learning) , cosmetics , ingredient , irritation , sensory system , context (archaeology) , artificial intelligence , gradient boosting , computer science , skin irritation , machine learning , medicine , algorithm , dermatology , biology , random forest , pathology , paleontology , neuroscience , immunology
Background The worldwide prevalence of “sensitive skin” group is estimated at being close to 40%. The main trigger for sensitive skin is the misuse of cosmetics products. Majority of the in vitro studies on cosmetic ingredients developed for topical application to the skin are not able to describe sensory irritation, such as stinging, burning, itching, and tingling. Besides, most of the in vivo tests often encounter problems such as limited number of subjects and usage scenarios deviate from reality. Objective A gradient boosting algorithm is adopted in our context to integrate multisource of information including skin types, sensory response, and cosmetics ingredients to predict sensory irritation. Method In this study, online comments were preprocessed to construct a multi‐dimensional structured data including skin types, sensory response, and cosmetics ingredients. A gradient boosting regression model was developed where sensory response was predicted based on the abovementioned structured input. The predictions were validated by in vivo test and were shown favorably when comparing with the state‐of‐the‐art results from related works. Result 46 007 samples were collected after data cleaning and were used in model developing. Results showed that the model was capable to predict the sensory response of 16 skin types to different ingredients (R = 0.71, P  < 10 −10 ). In addition, this model was validated using data from in vivo studies and presented a value of 75% on specificity, 88.9% on sensitivity, and 82.4% on accuracy. Conclusion Our approach that is based on a variant of the gradient boosting algorithm provided an adequate solution for understanding the sensory irritation of cosmetic ingredients.

This content is not available in your region!

Continue researching here.

Having issues? You can contact us here