Recommendation Algorithm And System Implementation Based On Collaborative Filtering

In the 21 st century,with the popularity of the network,the recommendation system has been greatly developed.Before the Internet is popular,people want to get a solution to the problem,they need to buy related books or consult relevant professionals.After the emergence of the Internet,more and more information on the Internet,the answer will also increase,people can't get the correct answer in line with their needs in time,you need to manually screen in the answer.The content or information provider wants to better serve the user,increase the user's stickiness,and hope that the user can quickly get their own demand information in their system,so the recommendation system will be generated,and its role will gradually appear,making many organizations And scholars have done more research and improvement on the recommendation system.Today's recommendation system has been extended to many fields,such as news,music,film and television,merchandise,etc.,which need to be personalized recommendations for users.The collaborative filtering recommendation algorithm has developed rapidly with its excellent performance.Collaborative filtering recommendation only pays attention to the relationship between users or projects,and does not need to pay attention to the content of users or projects themselves,so it is adaptable and can generate new recommendations and personalized recommendations.However,due to its wide application,it has exposed the defects of its system,such as system sparsity,low scalability,etc.When users and projects increase sharply,the user project matrix will be extremely sparse,making calculation accuracy.The decline,and the computing time of the system will increase,resulting in the system can't be recommended in real time,affecting its recommendation efficiency.This paper mainly aims at improving the similarity calculation method and data sparsity of the system.The research work mainly includes:(1)Since the rating of the project is evaluated by the user,the subjective emotion of the user will cause the scores of different users to become polarized,making their overall score tend to be high or low,resulting in an assessment due to subjective emotions.Standard different issues.Due to the extremely sparse data,it is found that the number of users' evaluations of movies is different,and a few users have inaccurate calculations because they have less common scores with the target users.Therefore,this paper proposes a scoring standard punishment mechanism and a low common score item punishment mechanism,using users.The score difference is used to constrain the scoring standard.The Tanimoto coefficient is used to calculate the proportion of the user's common scores.The above two correction schemes are combined.After several experiments and data fitting,a new similarity calculation formula is constructed,and a good recommendation is obtained.(2)Due to the increase of the amount of data,the sparseness level of the original matrix is gradually increased,which affects the performance of the recommendation system to a large extent,making the recommendation result inaccurate.In order to better solve the problem of data sparsity,this paper proposes project confidence based on the theory of information entropy theory and the idea of information entropy.Because the number of evaluations of different projects is different,and the number of high-evaluation projects is more representative of user interest,project confidence can better measure user interest.The traditional cosine similarity is combined with the project confidence,and the initial matrix is pre-filled to solve the original matrix sparse problem.In response to the different scoring criteria proposed above,the Euclidean distance formula is used to measure the relationship between users,and the recommendation is combined with the traditional algorithm.Compared with the traditional algorithm,the algorithm is superior to other comparison algorithms.(3)After the above research on collaborative filtering recommendation algorithm,this paper constructs a simple movie recommendation system,and presents the research algorithm in a systematic way.Finally,further research directions and ideas are proposed to solve the problems of the recommendation algorithm in a deeper level.