Local Government Debt Risk Rating And Early Warning Research Based On GBDT

With the development of China's market economy,local government debt on a large scale to develop local economy,but threatened by debt risk at the same time.Although the local government debt can effectively compensate for the lack of financial resources of local government,add the local government development and construction funds and provide the people's public expenditure with sufficient security,once the accumulated debt risk exceeds the local fiscal capacity,the credit chain will be broken,the regional financial risks will be out break.Huge debt not only makes the local government into financial difficulties and debt risk,but also inevitably into regional and systemic financial risks.In order to advance guard against this risk,it is necessary to predict the local government debt risk,timely keep local government debt and debt risk factors that lead to the local government.It is very important for the long-term stability of local government finance by avoiding in time.On the basis of the research on the construction of the local government debt risk comprehensive evaluation index,there is a subjective.judgment,which is lack of objective evaluation of the local government debt risk.TOPSIS(Technique for Order Preference by Similarity to an Ideal Solution)method is a kind of multi-objective decision making method,which is suitable for dealing with multi-objective decision problems.The principle is simple,can carry out multi object evaluation on each evaluation object and the pros and cons of sorting,fast computing speed,high resolution and objective evaluation results,has good rationality and applicability,high practical value.In the early warning method of choice,taking into account the risk data of China's local government debt transparency is not high,the difficulty of obtaining large,small sample,BP(Back Propagation)neural network model in the small sample training,the efficiency is not high,and the model itself into the local optimum and prone to overfitting disadvantages.GBDT(Gradient Boost Decision Tree)is a widely used algorithm,can be used for classification and regression,has good effect in a lot of data,and there are no defects of BP neural network,prediction method that is used as the local government debt risk is feasible.In this paper,the GBDT model as the local government debt risk early warning method is the first attempt in the field of GBDT algorithm.In this article,we set up a local government debt risk index system from three aspects,local government debt risk index,local government fiscal risk index and risk index of macro economy.Then,we use the TOPSIS entropy method to design the local government debt risk assessment index.According to the data of 31 provinces in China from 2014 to 2015,the training set and test set are divided randomly,using GBDT model for training of local government debt risk early warning model.Finally,compared with the BP neural network.According to the result of local government debt risk assessment index,the vast majority of local government debt risk in the provinces in the above medium alarm level.The improvement of local government debt risk is a universal phenomenon.From the results of the model,we can know the local government fiscal revenue and expenditure risk indicators can best explain the local government debt risk comprehensive evaluation index,and reasonable local government budget management can effectively prevent and control local government debt risk.Compared with the results of the BP neural network model,the GBDT model overcomes some defects of BP neural network,and its performance is better than that of the BP neural network.It is feasible and effective to make the GBDT model as the local government debt risk early warning model.