Risk Early Warning Of China's Shadow Bank Based On Wavelet Chaos RBF Neural Network

In recent years,with the continuous development and innovation of China's financial market and the increasing demand for social financing,shadow banking,as an effective supplement to commercial banks,has rapidly expanded and become an important part of the financial system.Although the emergence of shadow bank fills the demand of social capital,it also becomes an important source of financial risk.Therefore,it is of great significance to prevent the systematic risk caused by shadow bank.In this paper,the risk warning line and model of shadow bank are constructed to predict the risk situation of shadow bank and assist the relevant departments to formulate measures.On the basis of previous literature research,this paper selects 17 indicators from January 2011 to September 2019 from the dimensions of macro-economy,financial system and shadow banking subsystem to build China's shadow banking risk system.Using principal component analysis to reduce dimensions,six principal components are obtained,and according to the weighted average of the cumulative contribution rate of principal components,the comprehensive risk score of shadow bank is obtained,and the risk early warning line is established according to the criteria.Verify the effectiveness of the risk warning line in combination with major domestic financial events during the sample period.In this paper,we forecast the risk of shadow banks by comprehensive score.The comprehensive risk score is decomposed into three levels of wavelet,and the approximate term and detail term are obtained.The approximate term is the time series of denoising,and the detail term is the part of noise and chaos.Firstly,the chaos of the detail item is tested by the maximum Lyapunov exponent method.Three important parameters,embedding dimension,time delay and average trajectory period,are obtained by C-C algorithm and FFT transformation.The index is 0.0428,which proves that the detail item is chaotic.After that,the detail items are divided into training set and prediction set to reconstruct phase space respectively,and then they are learned and predicted by RBF neural network.At the same time,the approximate items are directly predicted by RBF neural network.At last,we reconstruct the forecast value of detail item and approximate item,that is,the forecast value of comprehensive risk score of shadow bank.In order to compare the prediction effect of wavelet chaos RBF neural network,the wavelet chaos BP neural network and ARIMA model are established and compared.Four indexes,mean square error root mean square error,mean absolute error and mean absolute error percentage,are selected to evaluate the model.The experimental results show that the prediction ability of wavelet chaotic RBF neural network is the best.