Leveraging Ingredient Profiles in Content-Based Skincare Product Recommendation

The fast expansion of the skincare sector has resulted in a bewildering assortment of skincare products available to users, making the choosing process challenging. This research describes the construction of an ingredient-based skincare product recommendation that helps users choose products that best fit their needs by using content-based filtering techniques with Term Frequency-Inverse Document Frequency (TF-IDF), cosine similarity, and Jaccard similarity. The research utilizes a dataset comprising various skincare products, including features such as "Product type", "Ingredients", and "Skin type". By applying content-based filtering techniques, this recommendation evaluates the similarity between products based on their ingredients, producing a ranked list of recommendations. Unlike common recommendations, which rely mostly on user reviews and ratings, this technique focuses on the specific ingredients included in skincare products. The proposed method obtained the highest accuracy of 80% for serum and sunscreen, and average accuracy for all categories was 66% using cosine similarity and 61% using Jaccard similarity. The proposed method is particularly useful for individuals looking for products with specific active components, as it emphasizes ingredient compatibility and efficacy over popular opinion.