On the Implicit Feedback Based Data Modeling Approaches for Recommendation Systems
Kübra TAŞ, Eyup Onder, Mehmet S. Aktaş · 2021 International Conference on Electrical, Communication, and Computer Engineering (ICECCE) · 2021
Online stores can give their customers access to millions of products, unlike retail stores. In online stores, with a large number of products to review, approaches that can recommend the right product to the customer are needed based on the customer's interests. In the literature, there are different methods that make product recommendations to customers, based on previous choices of customers. In general, we find that these studies are common when customers clearly provide ratings for their products. However, we observe a lack of data modeling approaches that indirectly capture customer ratings based on customer click data. For this purpose, within the framework of this research, we propose two different methods for indirect modeling of user ratings. We have developed a prototype application to show the feasibility of the proposed methods. The obtained data were tested with two state-of-the-art methods, alternating least squares (ALS) and collaborative filtering (CF). Clickstream data, obtained from an e-commerce book site, was used in the study. The results show that the proposed data modeling approaches can be used.