Comparing Recommendation Algorithms in Session-based E-commerce Sites

Recommender systems have become widely used in various website applications. With the integration of deep learning and recommender systems, the classic Session-based Recommender System (SRS) appears, which can obtain implicit feedback from explicit interactions. Some scholars have pro-posed many effective recommendation algorithms to provide better recommendation service in SRS. In order to compare the performance of these session-based recommendation algorithms, we consider a simple E-commerce SRS scenario and choose four representative session-based recommendation algorithms in this paper. Then we do some evaluation experiments. The experimental results show that the combination of local preferences and global preferences will improve the recommendation performance significantly, and for GNNs and RNN s in session recommendation based on deep learning, we also conclude that the prediction effect of GNN s is slightly superior to RNN s on long sessions.