多专长专家识别方法研究——以大数据领域为例

目的/意义]国家政府、大中型企业以及研究机构面对技术难题,如何找到合适的专家是迫切需要解决的问题。面对需要运用多学科知识来解决的综合性复杂难题,寻找到多专长专家显得尤为重要,寻找合适的方法识别出多专长专家是本研究的目的。方法/过程]利用专家所发表的学术论文数据,通过抽取专家有代表性的研究专长特征,基于TFIDF加权的重叠K-means聚类算法对专家进行重叠聚类划分,挖掘出专家的多个研究专长,进而识别出多专长专家。结果/结论]研究结果表明TFIDF加权的重叠K-means聚类算法在查准率、召回率和F值上有良好的表现,可以识别多专长专家。

Multi-expertise Researcher Identification: A Case Study of the Big Data

Purpose/significance]In response to the rapid shifting of knowledge needs, how to choose the appropriate researchers for a given problem is an important issue for the government, companies, as well as research institutions. When we face a real complex problem, it is essential to find multi-expertise researchers. This research aims to find a proper way to identify multi-expertise researchers. Method/process]This paper used a Term Frequency-Inverse Document Frequency (TFIDF) weighted overlapping K-means clustering method. Based on the researchers' co-authorship network built up from the publication data, the TFIDF weighted overlapping K-means clustering method was applied to cluster researchers into overlapping clusters and identify the multi-expertise researchers. Result/conclusion]Results show that the TFIDF weighted overlapping K-means method has an advantage over the previous work in terms of the precision ratio, the recall ratio and the F-value, so such a method can be beneficial to identify multi-expertise researchers.