<![CDATA[The selection of skincare products that do not match skin characteristics may cause problems such as irritation, acne, and dry skin. However, many users still experience difficulties in identifying their skin condition, resulting in less accurate product selection. This study aims to develop a skin-type classification and skincare product recommendation system based on the Random forest classifier algorithm by utilizing textual product data. The study employed a quantitative approach with an experimental method using secondary data from Kaggle, with ingredients and afterUse as the main attributes. The research stages included preprocessing, rule-based label construction, feature weighting using Term Frequency–Inverse Document Frequency (TF-IDF), an 80:20 data split, model training, evaluation, and implementation of a web-based system using Streamlit. The evaluation results showed that the model achieved an accuracy of 0.8229, weighted precision of 0.8215, weighted recall of 0.8229, and weighted F1-score of 0.8130. Therefore, the Random forest classifier is capable of supporting text-based skin-type classification with good performance and producing a web-based skincare product recommendation system.]]>
Copyrights © 2026
