The Recommendation System Improvements Under Online Privacy Protection

In recent years, the development of China’s Internet and e-commerce rapidly, dramatically changed the original study, work and live. Faced with a flood of information, recommendation system is used to solve individualized problems, such as shopping, mail and so on. But the attendant loss of privacy is increasingly grim situation. Information trading intensified, people’s lives and the economy have brought a great deal of damage, so the online privacy protection recommendation system is important and inevitable.In this paper, in order to protect privacy and improve recommendation systems, do the work of the following aspects:Firstly, the research of privacy protection and recommendation systems. By literature research of privacy protection, privacy impact and status, and recommend systems for indiscriminate use of the confidential information, I propose privacy protection policy solutions ideas, looking for privacy protection and recommend effect balance.Secondly, the user privacy concern quantified. In order to propose a solution, as this paper questionnaire survey approach, by exploring factor analysis to understand the user’s collection for privacy, cognitive, and control, the privacy information classified according to the degree of concern, and propose appropriate protection strategy.Thirdly, the propulsion system improvements under the Privacy Policy. After reading a lot of references and related research, based on the recommended system model, the first user data ETL information processing, and application-financial information K-anonymity algorithm privacy protection; then build a user model the user information and ratings data table cluster, replacing user with the concept of individual classes; lastly application recommendation algorithm similarity calculation, the TOP-N recommendation set.Fourth, the experimental verification system effects. This article will use Take-out ratings data to validate feasibility of privacy protection strategies and improve the effectiveness of the proposed recommendation system.