The Research Of Literature Ranking Prediction Algorithm

Academic ranking prediction means predicting the value of literature by certain algorithm and ranking the papers according to forecasts.In the academic social network,the academic ranking prediction algorithm can pick out those papers which will get widely intentions in future,and it is useful for them to grasp the current and future research directions.The most commonly used academic ranking criteria includes PageRank,HITS,etc.These traditional literature evaluation methods evaluate the literature by passing the authority through the citation.However,reference relationship implies the transfer of knowledge in citation network,namely the association between the citation and the cited literature,and the association has impact on the importance passing.And the citation network is just a snapshot of the current time,so it is unfair to the new paper.Meanwhile, in the typical paper ranking prediction algorithms,CiteRank algorithm only consider the paper’s publish time and FutureRank algorithm just involves the paper’s publish time,author’s authority and paper’s PageRank.Moreover,the relationship of the academic trends to be diversification,when the scholars retrieve papers,they will consider a variety of factors, such as the number of citations, similarity, influential journals/conferences and the authority of the author, these factors also implies an assessment of the value of literature. Therefore, when we predict the sort of literature,we should consider all kinds of features which may have impact on the literature value.To solve the above problems,A algorithm named FRP(Future Ranking Prediction), which measure the future influence of literature by predicting literature relative citations count to rank the academic paper, is proposed.The FRP algorithm involves four steps as follows:(1)based on the strength of association of the text between the citing and the paper which it cites and the cited time,we calculate the paper’s PageRank score and the Bonacich power;(2)we obtain the authors productivity,authority,average citation count and the co-author feature by combining the citation network and co-author network;(3)according to the paper’s publish time,we extract the citation rate in the citation network;(4)with features getting from(1)(2)(3),we use BP neural to predict the literature ranking.Experimental results show that our method improves the forecast accuracy.