Learning Sequential Behavior for Next-Item Prediction

抄録

A more precise recommendation plays an essential role in e-commerce. Representation learning has attracted many attentions in recommendation field for describing local item relationships. In this paper, we utilize the item embedding method to learn item representations and user representations. Our methods compute cosine similarity of user vector and recommended item vectors to achieve the goal of personalized ranking. Experiment on real-world dataset shows that our model outperforms baseline model especially when the number of the recommended item is relatively small.