Personalized Skincare Recommendations For Pregnant Women Based on Ingredient Safety And Skin Conditions Using Content-based Filtering

Abstract

Pregnancy brings significant physiological changes that affect skin health, with nearly 70% of pregnant women developing melasma and up to 90% experiencing hyperpigmentation. Despite the growing skincare market projected to reach USD 224.83 billion by 2034, pregnant women face challenges in identifying safe and effective products due to harmful ingredients such as retinoids, hydroquinone, and phthalates. Current solutions remain fragmented, requiring manual ingredient checks without integrated product recommendations. This research develops a comprehensive recommendation system for pregnant women’s skincare needs, using a two-component approach: (1) keyword-based classification to identify pregnancy-unsafe ingredients and categorize products by safety and therapeutic benefits for pregnancy-related skin conditions, and (2) content-based filtering with TF-IDF and cosine similarity to generate personalized safe alternatives. The dataset comprises 26.266 products scraped from the Skinsort platform, with evaluation on 1.042 unsafe products across eight categories. The quality performance evaluation showed cosine similarity scores ranging from 63.94% to 82.61%, with the highest in cleansers (82.61%), treatments (82.26%), and moisturizers (82.04%). The user evaluation using the ResQue framework, conducted with 10 participants, produced an overall mean score of 4.65 out of 5, categorized as “very high” across all dimensions. The highest ratings were obtained for perceived usefulness (4.8), attitudes (4.6), and behavioral intentions (4.78), reflecting the system’s effectiveness in supporting product selection. This research meets three objectives: implementing ingredient classification based on pregnancy safety and skin condition relevance, developing a content-based recommendation system using cosine similarity, and demonstrating effectiveness in meeting pregnant users’ skincare needs from both objective and user perspectives. The system offers a reliable, safe, and user-friendly solution enabling pregnant women to maintain effective skincare routines while ensuring safety, with strong potential for real-world application in supporting informed skincare decisions during pregnancy.