Personalized Recommendation Based On Probability Graph Model

Probabilistic graph model combines the knowledge of probability and graph,which is an effective method to deal with multivariable model,matrix decomposition in probability graph model is one of the most widely used and successful methods in personalized recommendation system.Probability matrix decomposition model for vast amounts of data processing and data score matrix data sparseness relief has a strong ability to solve,interest on the idea of this model is in a rating matrix was formed by scoring R,by means of dimension reduction and score matrix is decomposed into two implicit characteristic matrix U and V.The traditional probability matrix decomposition model ignores the inherent relationship between users and interest topics.Therefore,the Hawks model is used to analyze the correlation between users and interest topics,and define the relationship between user friends and interest topics based on this,this paper proposes a user interest forecasting model based on the social circle,and it also takes into account the interest and their own bias factor influence.Then we take into account the problem of interest drift,and then propose a comprehensive model based on the influence of timing factors and combining the influence relationship between users and interest topics.Finally,a comprehensive forecasting model of user interest is proposed,which is based on the influence of media,the stimulus of users and the decay of their own interest.