Researches On Event Recommender System Based On Probability Graph Model

In recent years,online shopping has become the main means of consumption,the number of users and consumer products continue to increase,leading to the Internet information overload.How users screen out their useful information from the increasing information base has become the focus of current research.The recommender system plays an important role in the research and plays an important role in many fields around people.From the initial recommendation of e-commerce items,it has been developed into the recommendation of news,short videos,events,activities and other fields.With the widespread application of recommender system in event recommendation,information overload and sparse score data can bring many problems to the accuracy of users' recommendation.To solve these problems,this paper will be based on the traditional recommender system model of the algorithm was improved,event recommender system based on probability graph model algorithm and validation data sets on its recommendation effect compared with the traditional algorithm has great improvement,the final figure will be based on probability model algorithm is used in the design of the recommender system a campus activities,showing a good application value.The main work is as follows:(1)Due to the problems of traditional matrix decomposition algorithms such as cold start,data sparsity and high prediction time complexity,the recommendation accuracy decreases.Normalized Matrix Factorization based on Implicit Feed Back and Baseline Predictor was proposed to improve the accuracy of user recommendations.The algorithm will give priority to data preprocessing,users and project evaluation matrix at the same time introduce Batch Norm sparse matrix algorithm to train the normalized processing parameters,to speed up the convergence speed,improve the stability of the training,at the same time increase the offset project constraints,more can show the user's true score,finally matrix to users to join the implicit feedback information,A matrix factorization model based on implicit feedback and benchmark preference is constructed to predict the user's rating matrix.Experimental results show that the proposed algorithm can solve the cold start and data sparsity problems in matrix factorization well,and has a great advantage over the traditional matrix factorization algorithm in prediction accuracy.(2)As some specific users do not like to score items after consumption,the data is sparse,leading to less reference data of the traditional recommendation model and unable to provide accurate recommendation information.a probabilistic matrix factorization algorithm based on specific user constraints is proposed.The algorithm introduces a potential similarity constraint matrix to influence the user's feature vectors for specific sparse users,and combines Maximum posterior probability estimation and Markov Chain Monte Carlo sampling inferred Probabilistic Matrix Factorization,automatically adjusted the regularization parameters of the model,and finally conducted test evaluation and comparison experiments on the data set.The experimental results show that the algorithm can greatly improve the prediction performance,and can accurately predict the user preferences in solving the problem of sparse rating for specific users.(3)Based on the advantages of matrix factorization model based on implicit feedback and benchmark preference and probability matrix factorization model based on specific user constraints,a campus activity recommendation system is designed and implemented by applying the model to practical life.The system can well recommend interesting activities to users,and generate more effective recommendation results by taking advantage of the activities users have participated in and the relationship between similar users,so as to meet user needs and increase user experience,Finally,the system is systematically tested to ensure the practicability and stability of the system.