Research On Algorithm Of Personalized Recommender Systems

Recently,recommender system is one of the foci in the field of big data science,relevant results have sprung up.In this paper,some algorithms and models related to the recommendation system are studied,especially the decision tree.Firstly,research background and status of recommender systems are introduced,some properties of recommender systems are displayed later,and then the evaluating system of recommendation systems.Secondly,the characteristic variables that will be used in the modeling of the recommendation system are analyzed,and two collaborative filtering algorithms based on user and object-based nearest neighbor recommendation are compared.A new splitting method is proposed,which is different from previous methods that split features step by step.The new method also aimed at getting maximum entropy,but split features after several steps.It's a method to split features several steps at once.It has put conditional distribution into consideration when making decision tree,so may get more information after splitting.Finally,a decision tree is trained by using click-to-rate data.There are two decision trees displayed with new splitting way.And it's more efficient when considering the node numbers,maximum levels and prediction accuracy.