Research On Online Privacy Preserving Based On Recommendation System Algorithm

With the advent of the information age,on one hand,the Internet has developed unprecedentedly and changed people’s daily life.On the other hand,information emerging exponentially every day,it is very difficult for users to quickly find out the products they are interested in the massive data.In order to select information effectively,the recommendation system has generated.The recommendation system is an available tool to solve the problem of “choice overload”,which could offer information users need and save their time and energy.Users’ information is a kind of resource,which inevitably leads to the issue of privacy disclosure.Especially in recent years,the privacy problems occur frequently,which has done great harm to people’s life and economic.Therefore,the privacy protection in recommended system becomes urgent.To protect privacy and improve recommendation systems,this paper mainly deals with the following three issues:First,the research on online privacy protection and recommendation system.This paper study privacy protection status quo,methods and the influence factors of privacy concerns throughout literature.In view of the situation of privacy protection ignored in recommendation system,this paper proposes privacy protection schemes,looking for the balance of privacy protection and recommend level.Second,privacy concern influence factors survey.To find the privacy protection strategies,this paper designs the questionnaire to know the actual situation of domestic online users,analyse the privacy concerns influence elements by factor analysis,including the user’s personal and privacy factors,perceived factors,credibility and compensation factors,and put forward corresponding protection advice.Finally,privacy preserving research of recommendation system algorithms.This paper study privacy preserving algorithm based singular value decomposition model by the interpolation of missing values and random perturbation techniques.On the basis of this,study privacy preserving method based on non-negative matrix factorization by adding effective auxiliary information.This paper tend to explore the balance between recommendation accuracy and privacy protection,verify the effect by the experimental data and use the index of the average absolute error and privacy level,analyze its effectiveness.