|
Application of The Bert Model for Classifying Types of Allergic Reactions in Consumer Reviews of Hypoallergenic Products |
|---|---|
| รหัสดีโอไอ | |
| Creator | Ungsumalin Sueawijit |
| Title | Application of The Bert Model for Classifying Types of Allergic Reactions in Consumer Reviews of Hypoallergenic Products |
| Publisher | MSMIS Thammasat University |
| Publication Year | 2569 |
| Journal Title | Journal of information systems in Business |
| Journal Vol. | 12 |
| Journal No. | 1 |
| Page no. | 123 |
| Keyword | BERT model, allergy classification, deep learning, hypoallergenic products, natural language processing (NLP), consumer reviews, text analysis, multi-label classification, skin allergic reactions, skincare cosmetics |
| URL Website | http://www.jisb.tbs.tu.ac.th |
| ISSN | 3088-1692 |
| Abstract | This research primarily aimed to apply the BERT (Bidirectional Encoder Representations from Transformers) model for classifying allergic reactions from consumer reviews of Hypoallergenic skincare products on the Sephora platform from 2018 to 2025. The main goal of this research was to develop a system capable of accurately identifying allergic reactions described in textual reviews, enabling consumers to make safer and more informed decisions when selecting skincare products. The research methodology began with collecting consumer review data from Sephora via the publicly available Sephora Products and Skincare Reviews Dataset on Kaggle. The dataset underwent a comprehensive data preprocessing stage to clean and standardize the textual data, followed by data labeling where allergic reactions were categorized using multi-label classification into various reaction groups such as redness, rash, acne, burning sensations, and other related symptoms. The next critical phase involved fine-tuning the BERT model, a powerful transformer-based language model renowned for its effectiveness in textual analysis. Finally, the model was rigorously tested and evaluated using key performance metrics including Accuracy, Precision, Recall, and F1-Score. The findings revealed that the fine-tuned BERT model effectively classified allergic reactions from consumer reviews with high accuracy, successfully identifying and categorizing consumer-reported symptoms associated with skincare product usage. Additionally, the model demonstrated the ability to analyze trend and extract associated risk factors from newer review datasets, highlighting its utility in commercial and public health applications. |
