Multi-view Transformation in Recommender Systems
Thi-Linh Ho, Anh-Cuong Le · 2021
In fact, there are grumpy users who tend to give lower ratings than others even if they feel well about the purchased items. This will lead to a decrease in the accuracy of recommender systems. In addition, users can rate purchased items from different views such as scores or comments. In this study, we proposed a model to transform the rating scores of grumpy users to match with other users by using users’ review text, then we used those ratings for improving the performance of the recommender systems. We did experiments on an Amazon dataset and compared the results of the proposed model with individual models of recommender systems to show that the transformation algorithm helped to balance users’ rating scores and changed the results recommended for users.