Research And Application Of Recommendation Based On Hybrid Filter

With the rapid development of information technology and the increasing content of information,the problem of information overload is becoming more and more serious,which brings great information burden to people.Recommender system becomes one of the most important ways to solve information overload problem with its characteristics at this stage.In particular,the rapid development of e-commerce has promoted the rapid development of the recommendation algorithm.The traditional recommendation algorithm,there are data sparse,cold start,development and other issues,and mixed current recommendation methods can effectively avoid some of the drawbacks of single algorithm,this paper proposes a hybrid recommendation algorithm can achieve effective features from the user interests,to achieve different demands on recommendation two aspects of user diversity and personalization.The development of deep learning that the recommended algorithm has ushered in new challenges,especially in predicting user commodity score,this paper introduces a design based on deep learning the latent factor model,which is trained by cascading noise auto encoder model,improved matrix decomposition training methods,improve the overall prediction training speed and accuracy;for to achieve the diversity and personalized recommendation results with consideration,this paper improves the related ideas of hybrid filtering,introducing the vector space model,realize the feature vector of the user description of goods,the final recommended by linear combination of mixed mode to realize the overall.In this paper,through the simulation experiment,the accuracy and diversity proved the superiority of the model method;on the other hand,the recommendation system function test,this paper proposes a hybrid recommendation algorithm can meet the demand of effective recommendation,to provide users with personalized and diversity both recommend.