Research On The Evaluation Of Chinese And American University Institutions Based On The Prediction Of Scholar Influence

The scientific research ability of university scholars is one of the main driving forces of national development and the basis of scientific and technological innovation.The starting point of this paper is to analyze the development of scientific research in China and the United States from the perspective of forecasting the influence of scholars and to evaluate the research potential of institutions.There are many research results on the evaluation of institutions,but most of them are based on the analysis of historical indicators.The development of scholars’influence in the future is closely related to the development of institutions,therefore this article has two goals:1)Build the feature set based on the time series characteristics of scholars’ alternative measurement,the time series characteristics of citation and the characteristics of personal basic information,and build the prediction model of scholars’ influence(H-index,citations per paper,citations per year and the number of papers published);2)Based on the research results of 1),this paper introduces the prediction index of scholars’ influence into the field of institution evaluation,builds an index model of institution evaluation,and further compares and analyzes the research potential of institutions in China and the United States.According to the research objectives,the research work is divided into four stages:(1)Data preparation and pretreatment,collect and sort out the basic personal data,citation data and alternative measurement data of computer science scholars in Chinese and American universities,preprocess and store the collected unstructured data,and carry out further statistical analysis.In this stage,a total of 2564 scholars from 30 universities in China and the United States were collected,and 241,070 pieces of citation and alternative measurement data were collected.In the process of statistical analysis,it is found that the number of computer science scholars in China is higher than that in the United States and there is a male-female imbalance among computer scientists in China and the United States.Computer science scholars in both countries are most widely distributed in the fields of computer science,engineering and mathematics,but the research of American university scholars is more dispersed than that of Chinese university scholars.Chinese scholars have advantages in the index of published papers,but they are slightly inferior to American scholars in the index of influence of papers.(2)Feature construction and feature screening,the features of time series data of different time Windows(Δt=5,Δt=3)are extracted,and XGBoost is used to filter the features to get the best feature set.The results of feature screening show that the timeseries characteristics of the total amount used and the total amount captured are very important to the prediction model of H-index and the prediction model of citations per paper.The time series of the total number of mentions,the total number of social media mentions and the number of doctoral degrees is all related to the prediction of citations per year and the number of papers published by scholars.(3)The prediction of scholars’ influence indicators,a variety of regression algorithms(Linear Regression,Support Vector Regression,Bayesian ridge regression,Gradient Boosting Regression,and XGBoost)were used to build the prediction model of scholars’ influence.R2 is taken as the evaluation index,and the optimal prediction model is selected.In this paper,it is found that the XGBoost scholar influence prediction model with a time window of 5 has the best performance,and models that include the altimetric features have better prediction results.(4)Establishment of institution evaluation model,in order to discuss the practical application of the prediction model of the index of scholar influence,the predicted value obtained from the model of the index of scholar influence is taken as one of the indexes,and the principal component analysis method is used to construct the evaluation model of the research potential of institutions from the perspective of scholars.At this stage,this paper extracted the five principal components to evaluate the institutions.The comparative analysis of the evaluation results of the mechanism evaluation model shows the availability of the model.However,the research in this paper can be improved,which will be supplemented in the follow-up research.