The multi-attribute ranking method based on Euclidean distance (TOPSIS-ED) can evaluate researcher academic influence through taking into account the different attributes of researchers. However, the method also has a problem that the points on the vertical line cannot be sorted. In this paper, we propose a multi-attribute ranking method based on relative entropy (TOPSIS-RE) by considering five indicators including the total number of papers, the total number of citations, citations per papers, I10-index and H-index. By calculating the relative entropy of the five indicators to the positive-ideal solution and the negative-ideal solution, this method ranks the authors according to the measurement that the results are close to the positive-ideal solution and far away from the negative-ideal solution. We select the American Physical Society data set as the training set and the authors who have won the Nobel Prize in the American Physical Society data set as the testing set. The area under curve (AUC) value is used to illustrate the accuracy of the algorithm. The results show that the AUC value calculated by TOPSIS-RE is 0.9321, and increases by 2.047% and 0.833% respectively compared with the total number of citations and TOPSIS-ED. Our work may shed some lights for quantifying the influence of scientists from the multi-attribute perspective.%基于欧氏距离的多属性排序方法(TOPSIS-ED)可以综合考虑科研人员的不同属性并对其影响力进行评价,然而该方法无法对其中垂线上的点进行排序.考虑作者的发表文章数、总引用量、平均被引用量、I10指数、H指数等5种指标,该文提出了一种基于相对熵的多属性排序方法(TOPSIS-RE).该方法通过计算作者的上述5种指标值与正理想解和负理想解的相对熵,根据其接近正理想解和远离负理想解的程度对作者进行排名.该文以美国物理学会APS数据集作为训练集,将获得诺贝尔奖的文章的作者作为测试数据集,用AUC值说明算法的准确性.实验结果表明,TOPSIS-RE方法算得的AUC值为0.9321,比总引用量指标提高了2.047%,并且比TOPSIS-ED方法提高了0.833%.该文的工作为从多属性角度刻画科学家影响力提供了借鉴.
展开▼
