| In scientific work,it is an important medium for storing and disseminating scientific achievements.In the age of the Internet,it is very,very easy for people working in scientific research to look up work related to a unique field.However,the number of papers is very vast,researchers need to evaluate the potential impact of the paper,so as to select the best download and study.One of the most commonly used metrics to evaluate the academic influence of a paper is the number of citations,which scientifically reflects the academic influence of a scientific paper from its publication to the present.The citation peak of most papers appears within five years after publication.Therefore,for papers have been published for a period of time,the academic influence can be evaluated by the citation times.However,for papers published for a short time,their citation times have not been fully accumulated,so it is difficult to predict the future potential and impact of these papers.Therefore,it is of great significance to predict the citations of academic papers.In this paper,we propose a novel method for predicting citations of a paper based on multi-dimensional paper feature information,including paper own feature,author feature,citation feature and journal feature.Two specific algorithms of swarm intelligence optimization,genetic algorithm and sparrow algorithm,are used to improve the optimization process and The traditional BP neural network was used to test the enhancement model using data from 750 papers in 29 enterprise management journals.The three-layer back propagation(BP)neural network model is selected for the undergraduate study based on practice analysis the RMSE and MAE improve by 18.16%and 22.7% respectively;For model optimized by sparrow algorithm,the RMSE and MAE improve by 25.78% and 22.81% respectively.The research conclusion is that BP neural network optimized by genetic algorithm and sparrow algorithm can be used to predict the citation times of papers,which can help researchers to evaluate and screen the paper quality when faced with a large number of papers and improve the efficiency of literature research. |
