Exploiting recurrent user activities for time-sensitive recommendation

Recommending sustainable products to the target users in a timely manner is the key driver for user purchases in online stores, which serves as the most effective means to engage the users into online purchases. However, most of the existing recommendation algorithms do not take into account the dynamics of recurrent user behaviors in recommendation processes. The two major but least explored challenges in this field are related to how to make the utmost desirable recommendation at the right time, and how to predict the user's next returning time to the service. This paper presents a novel method that combines self-excitation based on the Hawkes process and a collaborative filtering method based on the Temporal Matrix Factorization method to capture not only the temporal recurrent behaviors but also the change of users' interests that occur over time. Experimental results on various real-world datasets reveal that our model significantly performs better than all state-of-the-art methods