Rising Star Detection Based On Feature Analysis

For a long time,measure the academic value of a scholar has been studied in various fields from various perspectives by researchers.All academic institutions have also used this as the criterion for the distribution of research funds and scientific research awards.However,few scholars have devoted to studying how to predict the scholar's academic performance in the early stage of the academic career.In fact,for all academic organizations and various research institutions,in the increasingly large junior researchers,it's a significant and challenging human resource problem that accurately locate rising stars.The rising stars refer to junior researchers who may not be very outstanding in their research fields at initial stage of career,but tend to become expert scholars in their research fields and main members of their research institutions.However,when predicting the academic capacity of junior researchers,existing research programs in the selection of data as a basis for prediction is too rough,they are all random intercept 5 years as a research time window for processing,which is unreasonable.In the proposed model of this paper,when selecting research time window,we proceed as follows.We take the time factor of the scholar into account,regard the first paper as the start of his academic career,and intercept the data of the following 5 years as a basis of evaluation index.The proposed forecasting model adopts 7 characteristic factors as the basis of the evaluation: the number of papers,the total citation numbers,the average citation numbers,the number of collaborators,the average number of collaborators' citations,the node importance in the coauthor network and the node importance in the author citation network.Then we use machine learning to distinguish the weight of the above factors.Finally,we combine the weight and the factors to calculate the score of each scholar,and based on this score,we find rising stars.This thesis not only completes the construction of the prediction model,but also validates the simulation experiment on the actual dataset.In our experiment,the APS dataset in physics field and the DBLP dataset in computer field are used.The experimental results on the two datasets not only prove the validity and rationality of the proposed model,but also prove the universality of the prediction model.