TEAN: Timeliness enhanced attention network for session-based recommendation

Abstract Session-based recommendation task attracts more researchers’ attention in recent years. However, previous approaches suffer from limited timeliness since they overlook dynamic features of items and temporal semantic information, which results in inappropriate prediction. In this study, we propose an attention-based model named Timeliness Enhanced Attention Network (TEAN). It first extracts features of user and item from static and dynamic perspectives and then employs temporal semantic information by a time-cross mechanism. Our model is capable of ranking items based on timeliness enhanced features. Besides, we apply a pre-training method based on word2vec to learn embedding vector for users, items and temporal semantic information in an elegant way. Experiments on three datasets of different domains demonstrate that our approach improves performance opposed to other methods.