A Study On Local Government Performance Evaluation Based On Bayesian Network

Government Performance Evaluation as an important element in new public administration, government bears the responsibility to increase efficiency of government by all the attention. Since the 20th century mid-80, western countries in response to environmental conditions that scientific and technological progress, globalization and international competition, to resolve the financial deficit and public confidence issues, generally implemented the government performance evaluation measures that for the concept of customer first and strengthen government responsibility. The practice of western countries show that the implementation of government performance management and the government performance evaluation is an effective way to reinvent government, build a performance-oriented government and improve the implementation mechanism of the responsibility of the government. In today's information-based and knowledge-based era, under the influence democratization of public administration, economic globalization and market-oriented, government performance evaluation is a very timely topic for the government departments in many countries on the world.Bayesian network is firstly adopted in the evaluation of government performance and used to construct the evaluation model. Not only because it can systematically combine qualitative data and quantitative data, can be with graphics presentation, can clearly show the cause and effect relationship, and can support the inferences of incomplete and uncertainly data, but also it can deal with the problem of government performance evaluation that investigation is imperfect and too many subjective factors exist and so on.This dissertation firstly start from the theory and practice of the domestic and foreign government performance evaluation; Analyzes the connotation and significance of the government performance evaluation; Recalls practice of local government performance evaluation at home and abroad; Compares the qualitative and quantitative methods for local government's performance evaluation. Secondly, we construct local government performance evaluation index system from the point of public satisfaction adapted to our national situations; Thirdly, based on the field survey of the local government performance satisfaction, we construct Bayesian network model for local government performance evaluation and verify its validity and feasibility; Finally, this dissertation interpret the result of the model and analyze the impact of the index on the local government performance and provide countermeasures and suggestions.There are four main innovations in this dissertation: firstly, it creates an objective and effective "public judge official" government performance evaluation system from the perspective of an external evaluation by doing a survey of public satisfaction to overcome the defects of the "black box" nature of the past "official judge official" internal evaluation of government; secondly, based on related theory, learned from related practices, we build the local government performance evaluation index system in line with China's national conditions and specific conditions; thirdly, it introduces Bayesian network method to evaluate local government performance for the first time. This method can achieve multi-dimensional and multi-level analysis by model-based inference and overcome the shortcoming of other models that only remain in the index system construction and evaluation results; fourthly, explain the results according to the model and provide the corresponding countermeasures and suggestions.