Research On Recommendation Algorithm Based On Personalized Time Series Analysis

In the era of 5G technology and the continuous popularization of mobile Internet,the majority of users can obtain the information in the network more efficiently.At the same time,the Internet data is also growing explosively,which also causes the problem of information overload to some extent.Although the extremely rich Internet information can meet the personalized needs of a large number of users,it also makes it difficult for users to find the information they are interested in.Therefore,it is particularly important to filter out the redundant information that users are not interested in.At present,this kind of problem is mainly solved by building recommendation algorithms that meet the needs of users.However,the current research on recommendation algorithms also faces many challenges,such as user interest modeling,the influence of social networks on user interest,interest modeling integrating high-dimensional attribute information and so on.In order to cope with these challenges,it is necessary to introduce more valuable data to better model user interests and predict more accurately the projects that users may be interested in,and to dig deeper the potential information from the valuable data.Due to the more difficult to get user interaction-project contains records at the same time,users of social relations and project high-dimensional multi-dimensional comprehensive information such as attribute information,so this article will approach to the problem of decoupling,starting from the development rule of users' interests,dynamic changes of user interests modeling,dynamic characteristics of capture user interest,to recommend more in line with the current demand for the user.1.Development law of user interest.The interaction between the user and the project is the record that most directly reflects the characteristics of the user's interest,so this paper first from the user interaction-project data set to dig deeper into user interaction-project records in the potential relevance,build the user by means of sequence generated against network dynamic interest model,and through the interest model,based on user history for its predict user preferences based on time sequence of interest to the next project.2.Dynamic changes of users' interests.The user is a complex information fusion body,the demand for different information in life will be affected by the people around him,therefore,in order to explore this influence,this paper introduces social network information to explore the influence of social network on users' personalized serial recommendation.In this recommendation algorithm,the influence of the user's friends' interest characteristics on the user's interest is extracted from the user's social network graph through graph attention network.To build a dynamic interest model and generate recommendations for the next project of interest.3.Dynamic characteristics of users' interests.In daily life,users are interested in a certain project mainly because the project has the properties that make users interested.Therefore,users' interest characteristics can be constructed through the properties of the items that users are interested in.To explore the project properties influence on personalized recommendation,based on the fusion project high dimensional attribute information,and uses the collaborative knowledge map and attention to the network,between users-from a higher dimension mining project based on the properties of higher order connectedness,from a higher dimension to explore the characteristics of the user's interests to its formation has the interpretability of recommendations.