Research On Construction And Empirical Study Of New Models To Evaluate Author Academic Influence

Scientific communication plays an important role in scientific research.As a significant part of the scientific spread,it contributes to promote scientific development.In the system of scientific communication,authors are not only the core of it,but also the vital part of innovation power,and they conduct a crucial fundamental role in facilitating technological progress and cultural heritage.Generally speaking,authors who have great academic influence always drive more in knowledge innovation and dissemination.Based on the research status at home and abroad,we have done the relevant discussion first about the basic issues of author academic influence,including the definition of author academic influence from the perspective of citation,the influence factors of author academic influence,the relation between author academic influence and author academic level and now available methods of assessing author academic influence.And then we build new models to evaluate author academic influence from non-iteration and iteration perspectives.We weight the importance of authors according with citation in non-iteration.Also,we improve the weighted-PageRank based on the same principle.In the iteration network,author academic influence is related to direct citations as well as indirect citations.In addition,we takes international authors in informetrics as research objects and then analyzes and tests the differences between the two models and that with other four popular models—citation count,h-index,traditional weighted PageRank and weighted PageRank_D.In this progress,we take the ?Prize Award? as golden standard to measure evaluation effects of all methods,and choose some representative individuals to do deep analysis.Finally we also do a deeper and more detailed comparative analysis from the perspective of statistics,including ROC curve,AUC value,chart of frequency distribution,scatter diagram and Spearman correlation coefficient.The results show that the relation between the two models is high and significant,and nonetheless they are more or less different.The test effect of iteration model is dominant,while that of non-iteration model is ranked in the middle,but its AUC value comes to more than 0.9,indicating its test effect is also good.Non-iteration model is easily understood and conveniently operated and can distinguish the active scholars in the field;however,iteration model is more precise in terms of algorithm and can discern the foregoers.In practice,we can select proper evaluation model according to self-requirement.