Research On Measurement And Intelligent Early Warning Of Systemic Financial Risk In The Stock Market

Finance concerns overall development.Preventing and defusing financial risks has become a core requirement of the Party and the State for the development of the financial sector in recent years.General Secretary Xi Jinping clearly pointed out in “Several Major Issues in Current Economic Work” that both the symptoms and the root causes should be tackled,and a combination of long-term development and short-term development is needed to effectively prevent and defuse major economic and financial risks,and to firmly hold the bottom line of no systemic risks.Therefore,this dissertation focuses on the research theme of “systemic financial risk in the stock market” and deepens the research on the measurement and intelligent early warning of systemic financial risk in the stock market,which is of great economic significance in promoting the high-quality development of China’s capital market,preventing and defusing major risks,and firmly holding the bottom line of no systemic financial risks.In the study of systemic financial risk,measurement and early warning are complementary and organically connected,both of which are aimed at preventing and defusing systemic financial risks,preventing the occurrence of systemic financial risk events.Therefore,this dissertation combines measurement and early warning in the study of systemic financial risk in the stock market,and further deepens the study of systemic financial risk in the stock market.In the field of measuring the systemic financial risk of the stock market,first of all,timevarying generalized dynamic factor model is applied to measure the systemic financial risk among industries in the China’s stock market,and the early warning indicator of the systemic financial risk of the Chinese stock market is constructed based on the connectedness of the 2008 financial crisis and the 2015 stock market crash.It is found that(1)the impact of internal factors in financial system will significantly increase connectedness among industries,making it easier to cause systemic financial risks;(2)the connectedness of stock market has a time-varying effect that absorbs or may amplify market shocks for a long run;(3)Agriculture & Farming has the highest time-varying connectedness among industries in systemic risk events and it has the largest time-varying systemic risks spillover.Then,two-stage generalized dynamic factor model is theoretically further extended,and the systemic financial risk of China’s stock market is measured under the dual-cycle development pattern from the perspective of the international stock market.It is found that(1)a significant increase in the weak idiosyncratic component of international financial market returns and volatility was found in the 2015 stock market crash event,providing corresponding evidence for the financial contagion theory;(2)compared with time difference of market transaction,the level of capital market opening is an important factor that affects connectedness,interdependence and contagion between China’s financial market and international financial market;(3)Chinese investors has home bias effect,and the risk premium has the international investment preference;(4)common shock of volatility from the US stock market has the greatest impact on China’s stock market,and the prevention and control of the imported risk from the US financial market to the Chinese market should be focused on.Since the global financial crisis that erupted in 2008,domestic and foreign scholars have deeply recognized the important impact of abnormal asset price fluctuations on the stability of the global or regional economic and financial systems.A series of global and regional financial crisis events in the past have confirmed the interrelated relationship between asset price fluctuations and systemic financial risks Therefore,in the study of intelligent early warning of systemic financial risks in the stock market,first,the multi-lag neural network model based on wavelet transform and filter cycle decomposition(WT-FCD-MLGRU model)and the risk threshold model are combined to build a systemic financial risk intelligent early warning model with one-dimensional time-series.In the analysis of the effectiveness of the intelligent stock price forecasting model,the empirical analysis based on four global stock price indices(S&P500,IXIC,DJI,and SSE)shows that the WT-FCD-MLGRU model has a smaller forecasting error than the traditional econometric model(ARIMA model),machine learning model(SVR)and other deep learning models.In the analysis of the systemic financial risk early warning model,the classification indicators of the confusion matrix and risk warning thresholds are applied to test the early warning effect of the model on four stock price indices,and it is found that(1)based on the test of classification indicators of confusion matrix,the systemic financial risk early warning model constructed in this dissertation has better early warning effect on IXIC and SSE;(2)the systemic financial risk warning model can monitor abnormal fluctuations in asset prices to a certain extent and is effective in systemic financial risk warning by means of risk warning threshold test.Then,considering that the combination of technical analysis and fundamental analysis can effectively cover both non-systemic and systemic risks affecting stock price fluctuations and is the direction of stock price forecasting model development in the era of Fin Tech,this dissertation construct a new intelligent early warning model for systemic financial risks in the stock market by further extending the “one-dimensional time series” to“multi-dimensional time series”.In the analysis of the effectiveness of the intelligent stock price forecasting model,it is found through the empirical analysis of the stock prices of three commercial banks that compared with the previous research models,the hybrid recurrent neural network model based on the LSTM model and the GRU model can more effectively extract the data features of the technical analysis indicators and fundamental analysis indicators,thereby giving the optimal network model for forecasting the stock price.Moreover,the hybrid recurrent neural network model has higher prediction accuracy in long-term forecasting.In the analysis of the early warning model for systemic financial risk in the stock market,the classification indicators of the confusion matrix and the risk warning threshold are applied to test the early warning effect of the early warning model.It is found that(1)the early warning model has the best early warning effect on the positive and negative returns of BOC;(2)the false alarm rate of the systemic financial risk intelligent warning model with multivariate time series can reach the lowest 0.39% among the three commercial banks by risk warning threshold test;(3)the early warning effect of multivariate time series is lower than that of univariate time series,mainly because multivariate time series contain more information and are more complex.The application of multiple economic and financial indicators for early warning of risks is a future trend,so future research should further improve the early warning effect of the systemic financial risk intelligent warning model with multivariate time series.