Research And Application Of Paper Recommendation System Based On Citation Network Model

Due to the continuous development of scientific research,a large number of papers have appeared.This undoubtedly has brought problems to researchers when they are studying papers.These problems can be well resolved by paper recommendation system.Therefore,this paper proposes a graph model based on three layers citation network and recommendation algorithm.In the three layers citation network model,we combine the paper's content information and citation relationship,the author's coauthorship,as well as the research field information of the paper.The main research work of this paper such as follows:(1)In this paper,conducts certain amount of research in the recommendation system and the recommendation algorithm.Understanding the development status of researches.This paper summarizes the theoretical information of the recommendation system and citation network,and provides a theoretical basis for the model construction,algorithm application and the implementation of the recommendation system.(2)A three-layer citation network model has been constructed.At present,most citation network models only use a single feature and they don't involve the research field information.Therefore,integrate the citation relationship,paper's feature information,co-authorship relationship and research field information into this model.In this paper,adopts the DBLP-Citation-network data set on AMiner.Firstly extracted information about papers,authors and research fields from data sets.Secondly,established the paper citation network according to the citation relationship between papers,and established author co-authorship network according to the co-authorship between authors.Finally,according to the relationship between the three information,established the three layers citation network model.This three layers citation network model as a data model of algorithm and paper recommendation prototype system.(3)Put forward the algorithm of paper recommendation system based on the three layers citation network model: Combined the three layers citation network model and RWR algorithm to form the paper recommendation algorithm(PAFRWR).In this paper,chose recall@N and NDCG@N as the evaluation index.First,Contrasted with different evaluation indexes of ,and obtained the is 0.3.Secondly,through the experiment comparison of different search vectors,the evaluation index is higher when the search vector contains the triple information of paper,author and research field.Finally,compare with PageRank,LDA and Link-PLSA-LDA.The experimental results show that,the evaluation index value of PAFRWR is better than the other three methods.(4)Design and implement the paper recommendation prototype system based on the three layers citation network model: combined with this paper's model and algorithm,introduce the whole system frame and the design of the main modules.This system implementation by Python.And it with the XMLRPC model,convert the recommendation model into RPC service,and finally realize the system according to the technologies.