Research On Personalized Recommendation Based On Context-aware

With the fast development of Internet technology,the amount of information has exploded and greatly satisfied people's different needs for information.The massive data brings convenience to people,but it also causes serious problems of information overload.The recommendation system is a tool that can effectively overcome information overload.And it has been widely used in many fields.The recommendation system constructs the user preference model through analyzing the user's historical behavior records and digging the user's preference.Moreover,it actively recommends the most interesting content to the user to help users quickly find the information they want from the vast amount of information.It can also guide users to find out their potential preferences.The traditional recommendation system mainly makes use of the similar relationship between users and items to make recommendation,but seldom consider the situation information of users.Therefore,the recommended results cannot meet the needs of users in specific scenes,resulting in the low recommended accuracy and customer satisfaction.The context-aware recommendation system provides users with more accurate recommendations by incorporating situational information into the recommendation process,and it has become the focus and hotspot in the field of recommendation.This article focuses on the research of personalized recommend-ation methods based on context-aware.The main research work is as follows:(1)For the traditional personalized recommendation methods,most of them fail to consider the situation information of users or handle the situation information complicatedly.Based on the traditional latent factor model,this paper proposes a latent factor model with contextual information(C-LFM).The core idea of the model is to integrate contextual information into the latent factor model with generating an implicit contextual factor vector for each user and each item.In order to reduce the data dimensions and computational complexity,this model uses a supervised learning technique to extract potential contextual information and synthesizes various contexts into a factor of the model.While improving the recommendation accuracy,the model reasonably deals with the data dimension and computational complexity problems.(2)Considering the close relationship between user behavior and time factor and the importance of time context information to recommendation,based on the original random walk recommendation model,this paper proposes a random walk model with time attenuation factor(TA-RW),combining the time attenuation factor and the users' history rating records.The model fully considers the influence of time factor on the user's personal preference and the interest of social groups.And it integrates the time attenuation factor into the random walk model,and pays more attention to the recent behavior of the target user and the entire social group.The model not only accurately establishes the target user's recent interest preferences model,but also makes full use of social groups' recent preference information.In addition,the users' history rating records are introduced into the calculation process of user interest degree,and the calculation method of walking probability of random walk model is refined,which can better reflect the user's true interest in items.(3)Through a wide range of experiments and performance analysis,it is confirmed that the proposed methods have obvious advantages over the existing methods.