PRODUCT RECOMMENDATION SYSTEM USING IMPLICIT FEEDBACK BASED ON COLLABORATIVE FILTERING IN E-COMMERCE

Abstract

The high growth of e-commerce produces transaction data on a massive scale can be used as a marketing strategy by companies. One of strategy is a recommendation system that is used to predict interesting product information based on the characteristics of each user. However, recommendation systems generally use explicit feedback as a value of user interest in a product which creates a data limitation problem (cold-start) because only based on transaction data that has been rated by the user. Another solution could be using implicit feedback to avoid cold-start problems based on the number of user transactions for stores and product categories. In this study, the algorithm used is Singular Value Decomposition (SVD) to find similarities between one user and another based on the feedback value. The results of the model show good performance with score RMSE ± 0,865 and MAE ± 0,508.