Evaluating Non-Negative Matrix Factorization and Singular Value Decomposition for Skincare Recommendation Systems

Facial skincare plays a crucial role in maintaining clean, healthy, and radiant skin. Recommendation systems, such as Collaborative Filtering and Content-Based Filtering, can help users discover suitable skincare products based on their preferences and reviews. This study compares two Matrix Factorization techniques Non-Negative Matrix Factorization (NMF) and Singular Value Decomposition (SVD) to enhance the accuracy and relevance of skincare product recommendations. The results reveal that the SVD model outperforms NMF, achieving a Mean Absolute Error (MAE) of 0.7190, Root Mean Squared Error (RMSE) of 1.0104, Precision of 0.8054, Recall of 0.8144, and an F-1 score of 0.8099. In contrast, the NMF model produced an MAE of 0.7074, RMSE of 1.1052, Precision of 0.7865, Recall of 0.7987, and an F-1 score of 0.7926. These findings demonstrate that both models provide accurate recommendations, with SVD offering more precise and relevant predictions for skincare product recommendations.