A sequential recommendation algorithm combining attributes and long-term and short-term preferences
Yiyu Y. Yao, Yu Zhao, Liyun Zhao · IET conference proceedings. · 2023
In recent years, sequence recommendation has played an important role in solving information overload problems in many online services.Most of the existing sequential recommendation methods focus on the dynamic preferences in the time series of user interaction, and ignore the long-term preferences of users and the characteristics and attributes of users and projects, which limit the recommendation ability.To solve these problems, a novel sequential recommendation algorithm CASLRec is proposed in this paper, that incorporates both long-term and short-term preference and attribute information. This model embedded the user and the item into the vector respectively and added multi-head self-attention mechanism to extract the deeper connection between items from the order pattern of users. At the same time, considering the long-term and short-term preference of users with various weights assigned, it took the embedding of similar users as a supplement to enhance the user representation. At the same time, in order to improve the performance of recommendation, the embedded learning deeply integrates user and project attribute information to simulate user interaction more comprehensively. Experiment results show that the proposed algorithm on ML-100K, ML-1M, and ML-HetRec datasets is superior than other baseline models on HR@50 and MRR.