Neural collaborative filtering with multicriteria evaluation data
Hiroki Morise, Kyohei Atarashi, Satoshi Oyama, Masahito Kurihara · Applied Soft Computing · 2022
Recommendation systems help consumers find useful items of information given a large amount of information while avoiding information overload. Nowadays, in addition to traditional evaluation information (such as individual reviewer ratings), information such as multicriteria ratings are available on the Web. In the work reported here, we investigated whether collaborative filtering methods using multicriteria evaluation data and deep learning are effective for abundant and sparse multicriteria evaluation data. We also investigated whether adaptability can be achieved by predicting aggregated ratings from the evaluations of a few users. Experimental results show that three proposed methods using deep learning are better than traditional methods for both recommendation and rating aggregation.