Research On Predicting Scientific Elites Based On Growth Law

Scientific elites are the best researchers,and they promote social and national development.At present,the international community has gradually attached importance to the development of science and technology.Therefore,the number of scholars is increasing day by day.However,important scientific achievements are often discovered and created by a small number of researchers.Therefore,people are paying more and more attention to how to find scientific elites among a large number of researchers.Studying the growth process of researchers and discovering the law from this process is an important way to discover scientific elites early.In the process of growing into scientific elites,researchers interact with the external environment and material based on their innate qualities to form their own abilities,and finally gain social admiration through scientific creation activities.There are a number of common features in this process that can provide a basis for predicting scientific elites.Therefore,this article regards the members of the Chinese Academy of Sciences as representatives of scientific elites in China,studying the growth law of scientific elites from two aspects of personal and academic attribute characteristics,then extracting the corresponding indicators to construct scientific elites' prediction model.It provides the basis for scholar evaluation,talent training,and scientific research management.This paper is divided into two aspects:the study of the growth law of scientific elites and the construction of predictive models of scientific elites.First of all,the growth law of scientific elites is mainly studied from two aspects:On the one hand,the personal attributes of scholars:age,gender,birthplace,educational experience,work experience,administrative duties,and awards;On the other hand,the academic attributes of researchers:research cooperation,paper impact,personal impact.Finally,the researchers' growth rules are summarized:(1)Academicians are typical late bloomers with a long growth cycle;(2)The efforts of scientific elites can make up for the lack of quality education in the undergraduate stage.However,in the postgraduate stage,key universities,excellent research institutes and overseas study are indispensable platforms for their growth;(3)The working environment is relatively stable,and the number of job changes is negatively correlated with the growth cycle;(4)There is Matthew effect in their growth,65.3%of the researchers were awarded the national science and technology awards within five years before and after being elected as academicians;(5)More and more attention has been paid to the output and quality indicators of SCI papers,and there is little correlation between the output of SCI papers and whether they can become academicians.The combination of quality and quantity indicators is an important indicator to evaluate whether they can become academicians;(6)Scientific research cooperation is a common phenomenon in their growth process.When the number of co-authors is 6-8,the output and impact of SCI papers is the highest,and the ability of each scientific researcher in the team has been maximized.Secondly,this paper builds a predictive model of scientific elites.Through the study of the growth law of scientific elites represented by academicians,the index system of predictive models is initially constructed;then,through feature selection,it is found that the gender,cooperation,birthplace,the nature of supreme universities and sub-category administrative job characteristics are less correlated with the target variables.Therefore,13 indicators as input of models are obtained by eliminating and merging.Then,using three machine learning algorithms to construct the prediction model,through the prediction of the test set,it is found that the scientific elites'prediction model based on SVM algorithm is better than the other two models.Finally,in order to verify the application effect of the scientific elites' prediction models in real-life scenarios,using the constructed model to predict the academician candidates,the highest accuracy of the three prediction models is found to be more than 60%.The accuracy of the predictive model is within an acceptable range.