Research On Financial Crisis Crossing Contagion Based On Nonlinear Dynamic Model

In the context of economic globalization,the financial markets of various countries or regions are closely linked.Once the financial crisis breaks out,it will gradually spread outward through various channels,with a wide range of influence and far-reaching harm.If the existence of the financial crisis contagion can be proved,and the direction and intensity of the contagion can be judged in a timely manner,it will be of great significance for the country to manage and control risks and maintain national economic security in the face of financial crisis contagion.The paper focuses on the phenomenon of transnational contagion of financial crises,and introduces it from a nonlinear perspective.The nonlinear interdependence between dynamical systems is used to describe the contagion mechanism of financial crises between two countries through multiple channels.The nonlinear dynamic method is to treat the two countries as two nonlinear dynamic systems,applying the time series observations of different channel indicators to reconstruct the dynamic system through the time delay embedding method.Through improved nonlinear mutual prediction algorithm,the paper studies the dependencies between reconstructed systems,and then studies the phenomenon of the financial crisis.Compared with the original nonlinear prediction algorithm,the paper makes several improvements including extending the one-dimensional time series to highdimensional time series,introducing neural network model to train parameters,and modifying the expression of nonlinear prediction measure.The improved algorithm expands the indicators to three dimensions,comprehensively considering the indicators changes in the foreign exchange market,the currency market and the stock market.First the feedforward neural network is used in forecasting.Secondly considering that the variables are time series,long short-term memory network is introduced to improve the forecasting accuracy.Compared with the original measure,the modified nonlinear prediction measure can better reflect the prediction ability between systems.The above three improvements enrich the theoretical research on the contagion phenomenon of financial crisis based on nonlinear dependence,making the conclusion more convincing.The paper conducts an empirical study on the background of the US subprime mortgage crisis.The paper takes the UK and Russia,Japan and Thailand,and South Korea and Germany as examples to study the transnational contagion of financial crises.In the specific analysis,the improved nonlinear prediction algorithm is used to calculate the nonlinear prediction measure,and the contagion direction and intensity of the financial crisis in various countries are obtained.