Research On Dynamic Social Recommendation Algorithm Based On Combined Differential Privacy

With the rapid development of information technology,all kinds of social,audio and video,games and other software emerge as The Times require,and people’s life tends to be digitized and convenient.At present,there are some problems on the Internet,such as less available information,more redundant information,less real information and more false information,which seriously affect and reduce people’s quality of life.Although the current recommendation system has better solved these problems,it has greatly changed people’s life style and improved the quality of life.However,with the extensive and in-depth use of user data in the recommendation system,it is difficult to balance the prediction accuracy of the recommendation system with the security of user privacy information.The more the recommendation system knows about users and the more useful information they have,the better the effect of users using the recommendation system will be,and the more likely it is to leak user privacy information.At the same time,most of the existing recommendation algorithms are based on the prediction of the static recommendation system,that is,in a certain time interval that can be determined,the data of the recommendation system is used for calculation to achieve the purpose of recommendation.However,in real life,due to the continuous interactive use of netizens,users,items and scores are changing randomly,and the data used by the recommendation algorithm must also be in dynamic change.Therefore,the research of dynamic social recommendation algorithm based on privacy protection will be a research trend.Based on the above background,the research of this paper mainly does the following three works:Firstly,the social recommendation algorithm based on differential privacy and the social recommendation algorithm based on combined differential privacy in longitudinal data segmentation are studied,and the advantages and disadvantages of the two algorithms are compared and analyzed.Secondly,on the basis of in-depth analysis of the socialized recommendation algorithm based on combined differential privacy under longitudinal data segmentation,a dynamic socialized recommendation algorithm based on combined differential privacy under longitudinal data segmentation is proposed by introducing self-attention mechanism and time forgetting function,and the privacy analysis is carried out.Thirdly,the score data of multiple data parties are used to evaluate the social recommendation algorithm based on differential privacy,the social recommendation algorithm based on combined differential privacy in data longitudinal segmentation and the dynamic social recommendation algorithm based on combined differential privacy in data longitudinal segmentation.The privacy protection performance and prediction ability of the algorithm are compared and analyzed.