No abstract

Recommending relevant items of interest for a user is the main purpose of the recommendation system. In the past, those systems achieve the recommended list based on long-term user profiles. However, personal data privacy is becoming a big challenge recently. Thus, the recommendation system needs to reduce the dependence on user profiles while preserving high accuracy on the recommendation. Session-based recommendation is a recently proposed approach for the recommendation system to overcome the issue of user profiles dependency. The relevance of the problem is quite high and has triggered interest among researchers in observing the activities of users. It increased several proposals for session-based recommendation algorithms that aim to predict the next actions. In this paper, we would like to compare the performance of such algorithms by using various datasets and evaluation metrics. A deep learning approach named GRU4REC (Hidasi et al. in Session-based recommendations with recurrent neural networks, 2015) and simpler methods are included in our comparison. Real-world datasets from three different domains are included in our experiment. Our experiments reveal that in some cases of numerous unpopular items dataset, GRU4REC's performance is lower than expected. However, its performance is significantly increased after applying our proposed sampling method. Therefore, our obtained results suggested that there is still room for improving deep learning session-based recommendation algorithms.

No abstract