Collaborator Recommendation Research Based On Multi-dimensional Fusion

With the rapid development of the Internet and information technology,the traditional reference room and library are gradually transforming to the direction of informatization and digitalization,and the academic resource database arises at the historic moment.In the past 30 years or so,the academic resource databases such as CNKI,Weipu and Wanfang have been rising rapidly,and the scale of literature data collected and collated has been expanding continuously.The emergence of academic resource database has gradually changed the form of knowledge dissemination,greatly promoted knowledge sharing,and provided a platform for many scholars to read and display their academic achievements online.At the same time,rapid changes in science and technology and the explosion of academic information have made it difficult for scholars to find interesting and suitable partners to work with.For scholars,an accurate and appropriate recommendation result can promote research output,improve the quality of scientific research,and enhance users’experience and satisfaction with the recommendation system.Therefore,how to help scholars quickly find scientific research collaborators with higher academic level,similar research interests and higher cooperation probability in the vast amount of academic resources has become the key and difficult point for scientific research collaborators to recommend relevant research.This paper takes the academic resource information of CNKI and CSSCI as an example,and uses the methods of semantic analysis and social network analysis to mine the information from four dimensions of scholars’ natural attribute,interest attribute,ability attribute and social attribute.Finally,the recommendation of scientific research collaborators based on multi-dimensional fusion is realized with the help of expert rating and analytic hierarchy process(AHP).Among them,choose the academic nature age,title,or education background,subordinate to the province of three indicators,two indicators of academic age,title,or education background first rating,basis and be recommended after the similarity of scholars build scoring matrix,the province of institution index,according to the longitude and latitude to calculate distance,generated after provincial distance reciprocal matrix;The scholars’ interest attribute is embodied by the abstract of the scholars’ articles.The abstract is regarded as a text document,and a custom word list is constructed with the help of the article keywords.After word segmentation of the abstract information,the scholars’ text information topic mining and similarity calculation are carried out with the help of LDA topic analysis and cosine similarity method.Ability properties by the fusion of dispatch number year by year h index integrated embodiment and institutional authority,among them,merged the h index is considered the influence of scholars academic career,and considering the recent productivity,agency authority by the organization in the field of book intelligence discipline construction level of assessment and comprehensive rating"and reflected;The social attribute introduces the citation network and cooperative network of scholars,realizes the rating of the citation network of scholars with the help of Pagerank algorithm,generates the cooperative network of scholars and institutions respectively with the help of Gephi software,and calculates the point-centricity and mediation-centricity of scholars and institutions.In the end,11 indicators from 4 dimensions were normalized,and expert index weighting opinions were integrated with analytic hierarchy process to get the index weights.Then,the score of each index was integrated to get the recommendation results of scientific research collaborators.