Research Of Scientific Paper Recommendation Algorithm Based On User Preference And Citation Relationship

Due to the rapid development of Web2.0 technology,network information resource is increasing largely.While providing people with massive data,it also brought about the problem of “information overload”.In this situation,the recommendation system emerged as an effective information filtering technology to alleviate information overload,and has been applied in many fields successfully.However,the research and application of scientific paper recommendation is still relatively few.Recently,the explosive growth in the number of scientific papers has led researchers to spend a lot of unnecessary time to obtain the required papers,which hindered the effective advancement of scientific research.Therefore,it is of great value to research and implement a scientific paper recommendation algorithm which can provide accurate paper recommendation service.Through researching and analyzing the current relevant recommendation algorithms for scientific papers,and considering the characteristics of scientific papers,this paper proposes a scientific paper recommendation algorithm based on user preference and citation relationship.Based on matrix factorization model,the algorithm takes into account the user's implicit feedback information and the citation relationship between papers comprehensively to construct the partial sequence pairs and citation diagram respectively.And personalized paper recommendation is achieved by combining with the paper similarity.The comparison experiment proves that this algorithm can improve the recommendation accuracy effectively.Finally,a prototype system for scientific paper recommendation is designed and implemented based on this algorithm.The main work of this paper is as follows:(1)Reviewing the domestic and international research status on scientific paper recommendation system and analyzing the major problems it faces at present,so as to ascertain the research content and purpose of this paper.(2)Analyzing the commonly used recommendation algorithms in recommendation field and introducing the main models and technologies involved in this research briefly,thus establishing the theoretical basis for the research content of this paper.(3)A recommendation algorithm based on user preference is proposed,which using paragraph vector to calculate the paper similarity,and getting sorted papers by Bayesian probability model.Finally,the user's preference list is predicted based on the matrix factorization model.At the same time,the effectiveness of the algorithm is verified by comparison experiment.(4)Based on user preference,the citation relationship between papers is added,then a scientific paper recommendation algorithm based on user preference and citation relationship is proposed.The citation network diagram is constructed according to citation relationship,and the citation degree is calculated by the probability matrix factorization model,then the final scientific paper recommendation list is predicted by combining with the algorithm model based on user preference.At the same time,the effectiveness of the algorithm is verified by comparison experiment.(5)Based on the recommendation algorithm proposed in this paper,a scientific paper recommendation prototype system is designed and implemented.At the same time,the overall architecture design,main function module design and database design of the prototype system are described briefly.