Research On Sequential Recommendation Methods Of Time And Context Awareness

The main task of sequential recommendation is predicting users' next item based on users' historical events sequences and giving a personalized recommendation list.At present,sequential recommendation has been widely used in many fields such as point of interest recommendation,music recommendation and game recommendation.However,with the development of data collection technology in recent years,more data features have been applied to sequential recommendation,how to obtain effective information reasonably for recommendation has become one of the key challenges in constructing a sequential recommender system.First,the user's behavioral preferences may change dynamically over time.Secondly,there may be a negative effect between the user's behavior.For example,similar items that have been purchased may not be repeatedly purchased in the short term.At the same time,long behavior sequences may cause the model to lose some important historical information.Traditional sequential recommendations methods are difficult to solve the above problems and challenges effectively.This paper conducts research on the challenges mentioned above.Specifically,it focuses on analyzing analyzes the important feature between users' interaction events in depth,and proposes two methods for sequential recommendation.The main work of this paper is show as follows:(1)We propose a Hawkes process based sequential recommendation method to model users' preference as well as the complex correlation between different events.Specifically,the proposed method is able to mine the internal relationship between events,including positive and negative effects,which enable it to improve recommendation performance.The experimental results on real-world dataset show that the proposed model has better recommendation performance than the existing state of the art baselines.(2)In order to make fully use of time information in the sequence and deal with the problem of sequence information loss in long sequences,this paper proposes a novel sequential recommendation model that combines time interval and duration information to model the user's long-term and short-term interest accurately.At the same time,the global context information of the sequence is merged in the input layer to better utilize the long-term storage of sequence information.In addition,coupling inputs and forgetting gates are introduced to further improve the efficiency of the proposed model.The experimental results show that the proposed method has better recommendation performance than existing state-of-the-art methods in the modeling of long sequences.