Research On Academic Innovation Evaluation Of Papers Based On Machine Learning

As a form of scientific research,the thesis plays an important role in academic exchange and scientific development.In order to grasp the most cutting-edge research trends,researchers need to find innovative papers from a large number of documents,and to refer to them.Evaluating the academic innovation of the paper helps to promote scientific research and scientific research institutions to engage in more cutting-edge research,thereby promoting the development of science.At the same time,the results of the academic innovation evaluation can guide the formulation of science and technology policies,job ratings and resource allocation.At present,the academic innovation evaluation methods of the thesis are mainly divided into two categories:qualitative evaluation method based on peer re'view and quantitative evaluation method based on bibliometrics.In the qualitative evaluation method,peer review has the disadvantages of strong subjectivity,long cycle and high cost.This method is inefficient for the evaluation of a large number of documents.The quantitative evaluation method mainly includes a single characteristic index evaluation method,an influence measurement innovation method.an index system evaluation method,an evaluation method based on the content of the paper,and a structural innovation evaluation index method,and each method has certain limitations.Machine learning is the product of the development of the information age.Through the learning and understanding of multi-dimensional feature data,it can realize the evaluation of academic innovation of the paper in an adaptive way,and it is expected to solve the defects in qualitative evaluation and quantitative evaluation.Therefore,using the machine learning method to evaluate the academic innovation of the paper has become a new research directionThis paper evaluates the academic innovation of the thesis based on machine learning methods.The main research contents include:(1)Extracting the characteristic indicators related to academic innovation and constructing the evaluation index system of innovation;(2)Exploring the correlation between individual characteristics and innovation in the indicator system,and analyzing the individual characteristics of academic innovation.Mechanism of action;(3)Construct and test the proposed machine learning model(multiple linear regression model,regression tree,random forest,neural network model),select the model that has best performance to evaluate the academic innovation of the paper;(4)Using the regression tree model and the random forest model to analyze the importance of features,remove the characteristics unrelated to academic innovation,and correct the original evaluation index system.The experimental results show that the academic innovation evaluation index system constructed in this paper is reasonable.Based on machine learning,the academic innovation ability of the paper can be evaluated quickly.Among the selected models,the neural network model has the best evaluation effect.Through the analysis of the importance of the feature,it is found that the five characteristics of the co-factor and fund level in the original indicator system have nothing to do with academic innovation,and the index system is corrected based on this.The research in this paper provides a new perspective for the evaluation of academic innovation.The importance of the evaluation index system and characteristics of academic innovation can provide guidance and direction for subsequent research.