Application of Content-Based Filtering for Moisturizer Recommendation System Based on Skin Type Suitability

Abstract

Many users face significant challenges when trying to select the most suitable moisturizer for their skin. This difficulty often arises due to the overwhelming variety of available products on the market, combined with a lack of personalized information that could guide users toward the best choice. To address this issue, the present study aims to develop a recommendation system based on the Content-Based Filtering approach, which is specifically designed to align the benefits of moisturizer products with the unique needs of users' skin types. The data for this study were collected manually from 17 moisturizer products featured on the Sociolla e-commerce platform. Each product was carefully analyzed according to the descriptive information provided, including the benefits claimed and the skin types for which the product is recommended. The methodology involved several important steps: preprocessing the text from product descriptions, applying TF-IDF to assign term weights, and calculating cosine similarity scores between the user’s skin profile and product attributes. The analysis revealed that products such as D10 and D6, which yielded the highest similarity values, are strongly aligned with particular skin types. The resulting system demonstrates its ability to generate relevant and personalized product suggestions without the need for prior user preference data. This study concludes that using descriptive content as the foundation for recommendation logic can significantly enhance accuracy and targeting. Future enhancements may involve expanding the product database, integrating user-generated reviews, and leveraging machine learning techniques to produce even more adaptive and intelligent recommendations.