Research On Collaborative Filtering Algorithm Based On Similarity Calculation In Recommendation System

With the rapid development of information technology,the scale of Internet users and the amount of information are expanding.Network space information producers are transforming from the traditional media incline to the mass media and platform gradually.Along with the promotion of personal influence on network and the popularization of the artificial intelligence,everyone is the producer of information and the connected objects are the maker of information in the network space.For information consumers,the large amount of information is the sufficient information provider,at the same time also causes considerable distress,namely how to obtain the information from the cumbersome data space on internet particularly,the phenomenon is called information overload.The emergence of recommendation system can alleviate the overload problem effectively.The collaborative filtering recommendation technology can make use of the relevant information between users to produce reasonable recommendation.And it can be adjusted dynamically by the feedback information.But at the same time,collaborative filtering is also facing some problems,due to the increasing of user size and the scale of the projects,there will be the dimension disaster problem inevitably.The specific problems will be reflected in the sparseness of data.The effective using of the existing data information is the key to solve the sparseness problems in recommendation system.In view of the sparse problem,this thesis starts from the similarity research.First,based on the score attribute of the user,the user's interest distribution is obtained,and the data information is quantized based on the characteristics of the user's interest distribution.By calculating the difference of interest characteristics of the users,the similarity is obtained after the quantitative relationship adjustment.And the predicted score is carried out under the cooperative framework finally.Secondly,according to the membership characteristics of the user's score structure and the difference embodied in the user's scoring system,the similarity calculation method with the characteristics of user score structure and user scarcity is obtained respectively.In the limited user data space,a new similarity calculation method is formed through the two aspects of the characteristics with effective integration.Finally,the similarity method based on the interest distribution and the score structure can be used to obtain a similar relationship between users.The experimental results show that the proposed algorithm can alleviate the sparse situation and reduce the recommended error.