A User-based Normalization Multi-Criteria Rating Approach for Hotel Recommendation System

A User-based Normalization Multi-Criteria Rating Approach for Hotel Recommendation System

연구 분야: Verification

학회: SIET '21: Proceedings of the 6th International Conference on Sustainable Information Engineering and Technology

This paper develops a method that implements a proposed user-based normalization multi-criteria rating approach for the hotel recommendation system. The inspirations come from two main reasons. First, data normalization is a common problem in the multi-criteria based recommendation system. It occurs since users have diverse tendencies when giving ratings and that each criterion might have a varied range of ratings. Second, a previous study showed the superiority of the Decoupling technique for solving the normalization problem in a user-based multi-criteria recommendation system. In the meantime, researchers of non-recommendation systems showed that the MinMax technique dominates others – the Decoupling technique was not used in the study. These two facts raise a research question on whether MinMax could also outperform Decoupling when implemented on a user-based multi-criteria recommendation system. Therefore, in this paper, we propose implementing the MinMax normalization technique on a user-based multi-criteria recommendation approach and then comparing the performance to that of the Decoupling. Through series of experiments using the Yelp Hotel Dataset, we confirm that the MinMax technique can significantly improve the quality of a user-based multi-criteria hotel recommendation system better than the Decoupling. That is, the performance of MinMax is more than two times higher than Decoupling.