Research On Personalized Recommendation Method By Incorporating Contextual Information

With the rapid development of mobile Internet,wireless sensor technology and widespread popularity of intelligent terminal equipment,data is growing at an incredible speed.For the majority of users,they are faced with the problem of how to find the information they are really interested in from the massive data.Personalized recommender system is the most efficient way to solve the problem of information overload so far.Personalized recommender system predicts users' preference by means of their historical data,and recommends the information to users that satisfy their individual preferences.Traditional personalized recommendation technology ignores the influence of the contextual information of users on the recommendation results.But in real life,the contextual information(time,location etc.)has a huge impact on the user's choice behavior.For example,users usually may not choose to buy a T-shirt in winter,and they usually prefer to eat at a nearby restaurant.Obviously,we could improve the accuracy of personalized recommendation greatly by incorporating extra contextual information.This paper conducts an in-depth research on personalized recommendation by incorporating contextual information.The main research results are as follows:(1)This paper proposes a time-sensitive personalized recommendation method based on probabilistic matrix factorization technique.Firstly,we define and build a user-context matrix;then we propose a new time-dependence user similarity measurement method that based on user-context matrix by mining the implicit relationship between users in the user-context matrix.On this basis,a user similarity matrix based on context-aware is constructed;finally,the user similarity matrix based on context-aware is incorporated in probabilistic matrix factorization model.Experimental results show that this method performs significantly better than the state-of-the-art methods.(2)This paper proposes a novel multi-channel recommendation method with interpretability.Firstly,we introduce the user-based recommended explanation method and its measurement method of users' preference for items.On this basis,we propose a time-aware recommendation explanation method and a type-aware recommendation explanation method and their corresponding measurement methods of users' preference for items.Finally,we propose the multi-channel recommendation method with interpretability.The proposed method combine the recommendation results generated by traditional user-based recommendation explanation method with the results generated by the time-aware recommendation explanation method and the type-aware recommendation explanation method proposed in this paper.The proposed method puts emphasis on interpretability,and thus it can help to improve user's satisfaction and trust degree.Experimental results show that the method can guarantee the accuracy of recommendation.