Swarm Enhanced Attentive Mechanism for Sequential Recommendation

Recommendation system facilitates users promptly obtaining the information they need in this age of data explosion. Research on recommendation models have recognized the importance of integrating user historical behavior sequence into the model to alleviate the matrix sparsity. Although deep learning algorithm with attentive mechanism exhibits competitive performance in sequential recommendation, the searching for optimal attentive factors still lack effectiveness. In this work, we redesign the sequential recommendation model by employing swarm intelligence for optimization in the attentive mechanism thus to improve the algorithm accuracy. We conduct extensive comparative experiments to evaluate performance of four swarm intelligence algorithms and traditional recommendation methods. Our work is the first attempt to integrate swarm intelligence into sequential recommendation algorithm. Experimental results confirmed the superior performance on AUC score of the proposed approach.