Long Short-Term Memory with Sequence Completion for Cross-Domain Sequential Recommendation

As the emerging topic to solve the loss of time dimension information, sequential recommender systems (SRSs) has attracted increasing attention in recent years. Although SRSs can model the sequential user behaviors, the interactions between users and items, and the evolution of users’ preferences and item popularity over time, the challenging issues of data sparsity and cold start are beyond our control. The conventional solutions based on cross-domain recommendation aims to matrix completion by means of transferring explicit or implicit feedback from the auxiliary domain to the target domain. But most existing transfer methods can’t deal with temporal information. In this paper, we propose a Long Short-Term Memory with Sequence Completion (SCLSTM) model for cross-domain sequential recommendation. We first construct the sequence and supplement it in which two methods are proposed. The first method is to use the intrinsic features of users and items and the temporal features of user behaviors to establish similarity measure for sequence completion. Another method is to improve LSTM by building the connection between the output layer and the input layer of the next time step. Then we use LSTM to complete sequential recommendation. Experimental results on two real datasets extracted from Amazon transaction data demonstrate the superiority of our proposed models against other state-of-the-art methods.