The Application Research On Personalized Recommendation In Social Media Environment

Along with the network information resources continue to accumulate and expand,users become more and more difficult to collect their required data Recommendation system ease the problem of information overload,effectively mine information resources,and initiative push to the user.At present,the most widely used collaborative filtering recommendation algorithm is applied in the system.In recent years,social media is becoming increasingly popular,researchers can solve the problem of cold start by introducing trust relationship data in social media.In the absence of full user rating data,reference to the score of the trusted users to predict the score.The key of the algorithm is how to effectively use the data,so as to improve the recommendation.In this paper,I research the traditional user based collaborative filtering recommendation algorithm and the trust relationship based collaborative filtering recommendation algorithm in social media environment,and than put forward my own improved algorithm.The main research work of this paper as follows:(1)An improved collaborative filtering recommendation algorithm based on trust relationship is proposed,which is based on the deep mining and the effective use of the data.In view of the existing recommendation algorithm in trust intensity evaluation model is not perfect,the potential trust intensity in the social trust network data and user rating data is mined and synthetic to the original trust intensity evaluation model,so that the model can obtain a new trust intensity.The potential trust intensity is expressed in the common trust user ratio of the social trust network data and the common score ratio of the user score data.In the last,the new trust intensity is applied to the prediction model based on trust relationship,so as to get more accurate result.(2)In order to further improve the accuracy of the recommendation,overcome the shortcomings of the general hybrid recommendation algorithm,which is a static hybrid recommendation algorithm.A dynamic feedback collaborative filtering recommendation algorithm based on the fusion of similar users and trust relationship is proposed.The algorithm uses dynamic factors to fuse the similarity between users and trust relationship.Dynamic factor sets random value at initialization time.A positive and negative feedback mechanism is established according to the error of user feedback and system prediction.In accordance with the feedback type,select the value added or attenuation function properly adjust the dynamic factor,so that the system can better predict the user's score.(3)Using the real data set Epinions,the proposed algorithm is experimentally verified.The experiment include experiment design and experiment results analysis.By comparing the experiment results,we can see the improved algorithm based on trust relationship can improve the accuracy of recommendation,and it will not lead to damage the coverage ratio of the original recommendation algorithm.The fusion algorithm of dynamic negative feedback overcomes the static of the general mixed recommendation algorithm,and further improve the recommendation accuracy compared with the recommendation based on similar users or trust relationships,and guarantee the recommendation coverage ratio.