Research On Personalized Recommendation Algorithm Based On User Behavior In Big Data Environment

With the rapid development of information technology,it has covered into every corner of society,economy and life.The information about the relevant activities in various fields is recorded as data preservation.Advances in technology have substantially reduced the cost of data storage,massive data are stored in databases or in the cloud,the amount of data is growing at an unprecedented rate,and gradually push us into the era of big data.In the context of the era of big data,behind these huge data is often hidden amazing value.However,the data in big data environment presents the characteristics of scale,diversity,real-time,low value density and so on,which greatly increased the difficulty of analysis of data mining,so that the utilization of data is far less than its growth rate,resulting in more and more serious problem of information overload.At present,personalized recommendation system is one of the effective ways to solve the problem of information overload,so the recommendation algorithm in the recommendation system has become one of the most popular research directions.In this dissertation,based on the background of big data for the background,optimize the recommendation algorithm as the main research goal,aiming at the key problems of the recommendation algorithm to improve,it is expected to improve the accuracy,the diversity and novelty of the recommendation algorithm.In this dissertation,the following theoretical and practical research:(1)In order to solve the problem that the collaborative filtering algorithm of Restricted Boltzmann machine(RBM)is easy to assimilate the user's personalized demand in the prediction stage,this dissertation proposes a collaborative filtering recommendation algorithm based on the nearest neighbor.There is a high degree of similarity between the interests of the nearest users,and the users of the same interest are score closer on the same item.According to this view,calculating the probability of the grade of a project score which the user does not score and the nearest neighbor scores.The probability is incorporated into the RBM model prediction stage to strengthen the user's personality in the prediction results and improve the accuracy of the algorithm.The experimental results show that the improved algorithm not only improves the accuracy of the algorithm,but also enhances the ability to resist over-fitting.(2)In order to solve the problem that the collaborative filtering algorithm of RBM prediction accuracy of the user who has a different opinion of "popular project" is poor and the discrimination of prediction "unpopular project" is poor,proposes a collaborative filtering algorithm based on item tag for RBM.Use the objective tag of the project itself(such as the theme of the film,the commodity category etc.)describe user's preference,this procedure takes advantage of the user's own project information which rated projects,strengthen the individual needs of users.And the prediction of the "unpopular project" is more objective and true,and the accuracy of prediction results is higher.Finally,the experimental results show that the prediction accuracy of the algorithm is improved by 1.2%.(3)In order to solve the problem that recommendation method based on network structure excessive recommended "popular resources",ignoring the recommended "unpopular resources",propose a new algorithm based on weighted network structure.In this dissertation,the method of energy diffusion in the network structure is improved.In this dissertation,we propose a new algorithm based on weighted network structure.In this dissertation,we improve the way of energy diffusion in the network structure,to improve the recommendation for the "unpopular resources",and to improve the diversity and novelty of the recommendation results.The experimental results show that the proposed algorithm not only takes into account both the accuracy of the recommendation,but also effectively improves the recommendation of "unpopular resources".