Skincare Recommender System Using Neural Collaborative Filtering with Implicit Rating

Skincare products are essential cosmetics for women, especially in this modern era. Many e-commerce services provide a variety of skincare products in their catalogs. One problem with purchasing skincare products online is that users cannot try the product and depend on other customers' rating reviews. However, rating reviews on a scale of 1 to 5 are considered insufficient to represent product quality, and users need to read review texts written by other users to get more specific information about the quality of the product. This paper investigated NCF (Neural Collaborative Filtering) for skincare recommender systems. Instead of using explicit rating as usually used on standard recommender systems, we adapted the sentiment score as a rating which, in our experiments, proved can improve the classifier's performance. We collected 180,104 rows of data with 11 data attributes and 1,339 skincare products to evaluate our proposed method. Experiments on the dataset show that the proposed NCF with explicit ratings achieved an RMSE of 0.8033, and the NCF with implicit ratings achieved an RMSE of 0.4931.