Research On Tag And Privacy Protection Based Cluster Recommendation Algorithms And Their Application

As an important tool of information retrieval,recommendation system has been widely used and developed in many fields such as e-commerce.With the introduction of the concept of web2.0 in 2004,the Internet has entered an era in which users can mark content freely.Social tags are widely used in the field of e-commerce.Users can better classify resources according to their preferences through free tagging.The recommendation system can also predict users' interests based on tags and make efficient recommendations.Recommendation algorithm based on user information,resources,information and browse the content such as preference prediction recommended,makes recommendation system is facing the following problems: the expansion of the amount of data to make it face data sparseness problem,according to the user's information and browsing history to predict when interest preference privacy problem,the change of user interests over time migration and so on.In view of the above situation,the corresponding research and improvement,the main content is as follows.(1)Research is conducted on data sparsity,and Slope one algorithm is proposed to fill in the score prediction of resources that have not been evaluated by users,so as to alleviate the problem of sparsity.Tag as a connecting bridge between users and resources,through score better response user attitudes towards resources when using the tag,so will score and fusion tags.It is more helpful to improve the accuracy of recommendation by weighting tags through scoring matrix.(2)The privacy disclosure problem is studied,and the method of introducing differential privacy protection idea in fuzzy c-means clustering is proposed.To join in the process of clustering Laplace noise,protect the clustering center and clustering center with privacy protection and with similar users privacy protection,to achieve the purpose of protecting privacy.(3)The initial center randomness problem of the fuzzy c-means clustering algorithm is studied,and the fusion method of density idea and maximum and minimum distance idea is proposed to solve the problem of membership degree matrix and the initialization randomness of the clustering center.To have a higher accuracy and faster convergence speed.(4)Aiming at the problem of user interest change with time migration are studied,put forward the improved index lost function,with its tag weighted response user changes in short-term interest;In addition,the concept of time window was introduced to give consideration to each person's long-term hobbies for scoring prediction.To achieve dynamic simulation of user interest over time.(5)Use the above research results to design and implement the film recommendation system from the aspects of demand analysis,architecture design,module design,database design and so on.Thus,the clustering recommendation method based on tag and privacy protection is applied to the personalized movie recommendation system.Achieve the recommendation target of "customization,privacy and security" for different users.