Analysis And Application Of User Behavior Data Based On Stable Interest And Interest Evolutionary Model

The recommendation engine in various fields have been widely recognized.Many of the recommended engine considering the factors of user interest,but most recommendation engines only focus on the user's one interest.They did not integrate evolution interest with stable interest,and the prediction accuracy needs to be improved.Current recommendation engines are highly targeted at one domain,but lack of general considerations,so resulting in that recommendation algorithm fits for one domain well,but is difficult to apply to other areas.Based on in-depth analysis of user preference model on the traditional concept of interest and interest in the stable evolution of user preference model,the new user preference model of organic combination of the two interest is established,and further promoted as a general model.First of all,based on the personalized recommendation algorithm commonly used in recommendation system were summarized,the related theory and technology are introduced.Based on the traditional user interest models and the relevant theory of psychology,stable interest and interest evolution to proposed and applied to analysis and calculation of the user interest,and is put forward to general model of the evolution and stable interest.Association rules is used to solve the problem of sparse matrix and cold start problem.Acurate recommendation is achieved in a small amount of data.Through combining of the above factors and revising the parameters of the generic model,an optimilized interest recommendation model is made to meet specific environment with less user behavior.When new users visit application,background application recommended model will analyze characteristics of the user's environment and the user,and give the best recommendation.When the user has left its traces in the application,system background will use interest evolution model to correct for personal interest in stability,and infer the user's current interest according to the user traces.In this paper,data of nearly a year of user information related to a community maternal and child supplies APP's post,circle,log and user behavior knowledge is use to test out algorithm.After pretreatment,9823672 users' behavior data generated in the application,including the post,browse,copy,share,evaluation,feedback that knowledge,question and answer is given.Test results show that,compared with the traditional user interest models,the optimized the recommendation model and algorithm in this paper not only improved precision rate and recall rate,but also solved the problem of interest in sparse matrix.The reliability,high availability and adaptability of the algorithm of based on the interest model in general is verified.