Research On Evaluation Of Domestic Think Tank Based On BP Neural Network And MIV Algorithm

The degree of development of think tanks is becoming an important manifestation of the governance capacity of a country or region.Further considering about the ways and strategies to improve service decision-making of think tanks could be obtain by exploring the evaluation methods of think tanks and understanding the methods and indicator systems in think tank evaluation.At present,Chinese national high-end think tanks and provincial and municipal key think tanks are developing rapidly.The scientific and efficient evaluation system can encourage think tanks to continuously seek development,which is of great significance to the formation of Chinese new think tank system.Starting from the evaluation of think tanks,this paper demonstrates the advantages of using BP neural network to evaluate domestic think tanks.Then based on the literature,the characteristics of domestic think tanks and expert advices,a questionnaire is produced.The result of the questionnaire is analyzed by factor analysis,and the hierarchical relationship and weight of the indicators are clarified.On this basis,we establish a domestic think tank evaluation index system based on the principles of science and comprehensiveness.Then a domestic think tank evaluation model based on BP neural network and MIV feature screening is constructed.The model is based on Python 3.6,using the TensorFlow framework,JetBrains PyCharm and BP algorithms.The training and verifying sample data is obtained from 100 known-level domestic college think tanks,using its secondary indicator weighted scores and known ranks.After the establishment of the domestic think tank evaluation model based on BP neural network,in order to minimize the accuracy problem of the model caused by the inaccurate selection of evaluation indicators,the MIV algorithm is used to screen the correlation of neural network variables.Seven key indicators are selected from the eight secondary indicators as input layer nodes,and the comparison results are obtained from the perspective of training time and evaluation accuracy:the domestic think tank evaluation model based on the MIV algorithm optimization will slightly make the model training time longer,but it performs better in terms of evaluation accuracy.Therefore,this paper selects the domestic think tank evaluation model of seven key evaluation indicators.In the end,this paper takes the annual assessment of 16 key think tanks in Jiangsu Province in 2018 as an example.The predicted grade results are basically consistent with the actual assessment results of the example.This paper establishes a domestic think tank evaluation model based on BP neural network and MIV algorithm,and demonstrates operability,scientificity and practicability in the actual think tank evaluation work,enriching the research method of think tank evaluation,and providing new evaluation for practical think tank evaluation.Method reference.It is proved that the evaluation model optimized by the MIV algorithm is useful.This paper establishes a domestic think tank evaluation model based on BP neural network and MIV algorithn,and demonstrates operability,scientificity and practicability in the actual think tank evaluation work,which enriches the research methods of think tank evaluation,and provides new research method for practical think tank evaluation.