Research On Collaborabive Filtering Algorithm Based On Data Detect And User Trust

With the gradual increase of the information resources on the internet, users'demands for personalized services are increasingly heightened. Personalized recommendation technology is an important research subject for e-commerce recommendation system. The collaborative filtering for the personalized recommendation is by far the most widely used and the most successful personalized recommender technology. While, with the recommend systems widely used, the recommendation algorithm's security problem is increasingly appear. As the recommendation algorithm can not effectively remove malicious data, resulting in recommendation system can not effectively get user's real interests, so the recommendation is abviously deviation from the user's interest. How to ensure the quality of collaborative filtering algorithm has become a major problem. On the basis of comprehensive analysis for the research at home and abroad, this paper has further deep research on collaborative filtering recommender technology.Fistly, deeply analyzing the traditional collaborative filtering recommender algorithm, especially for the recommender system can not effectively resist the injection of malicious data, resulting in recommendation system recommended for serious errors; this paper proposes an improved collaborative filtering recommender algorithm with user trust value. This improved algorithm uses a new similarity method, taking into account the impact of user rating on items, by importing the credit to inhibit the malicious user data. The method can effectively reduce the negative impact on recommendation systems.Secondly, in view of the problem of malicious data existing in recommend system, this paper introduces a concept of character for item rating and proposes a kind of method to detect the malicious data by calculating the sum of the items's similarity value. With evaluation metod help to determine whether the item was affected by the malicious data.Finally, it also gives out the analysis and verification to all the technologies which are mentioned in this paper. Subsequently, it is compared with the existed typical algorithm and also makes the prospects for the future research.