The Impact Of Economic Uncertainty On The Quality Of Stock Index Monitoring Under The Composite Neural Network Architectur

The foundation of national security is economic security,and financial security is the key to economic security.The construction of financial risk early warning and monitoring system has been the focus and focus of continuous attention of academia and the industry.The monitoring of key stock indexes has been the primary process of the national securities regulatory authority since the great turbulence of the securities market in 2015,which has been included in its daily high-frequency routine work.At the end of September 2022,China quickly completed the establishment of a national-level financial stability fund to calm the sharp fluctuations in the capital market.At present,the "black swan" and "grey rhinoceros" dance together.The epidemic situation is normalized around the world.The risks of economic uncertainty are looming.The war of the Russian-Uzbekistan conflict has cast a heavy shadow on the recovery of the world economy.The "eagle" interest rate increase of the Federal Reserve has led to the tide of the US dollar,which has seriously impacted the exchange rate markets of various countries and triggered high intensity financial turbulence.Under the circumstances of such uncertainty and risk,it is particularly necessary and urgent to further improve the research of stock index monitoring.At present,the core stock index monitoring system adopted by the industry is mostly based on the rational multi-factor model and rarely incorporates the irrational factor.This study attempts to introduce the irrational factor into the stock index monitoring research,and conduct theoretical research in the perspective of behavioral finance.On the basis of the use of financial measurement and machine learning algorithms,it constructs a deep learning algorithm model,and attempts to establish a nested composite network model of non-single class neural network algorithms,It also introduced artificial intelligence to optimize response parameters,trying to achieve a more accurate,efficient and reasonable composite model,complete the missing modules of the existing stock index monitoring model,improve the financial risk monitoring and early warning system,and further strengthen the security guarantee for realizing the great rejuvenation of the Chinese nation.The stock index monitoring includes two modules: stock index prediction and causal back-test.After empirical test and theoretical research,this thesis believes that the irrational factors represented by economic uncertainty(EPU)have a cyclical,regular and critical impact on the level and volatility of the stock index,and closely affect the prediction and back-test of the stock index monitoring system.In the prediction module,the incorporation of EPU into the algorithm model can achieve a significant improvement in the horizontal improvement of the prediction error under the same algorithm compared with that under the non-Na model.After integrating the nested depth neural network model of the short-term and short-term memory network LSTM and the graph neural network GRU,and integrating the particle swarm optimization algorithm PSO,the goodness of fit under the single neural network model is significantly improved.After the financial orthogonal and financial click to prevent over-fitting,The month-level horizontal convergence of the infinite order experimental prediction error of PSO-LSTM-GRU composite neural network is about 30%,the longitudinal convergence is about 20%,the day-level horizontal convergence is about 45%,and the longitudinal convergence is about 30%.In the back test module,the more accurate the spatial dimension of factor returns is,the more efficient the fast positioning and cause finding function of the stock index abnormal early warning back test is.The contribution degree and validity period of rational factors will change over time.There is no perfect factor that shows positive or efficient performance at every stage and every moment.Regular and irregular "water change" operation of rational factor pool will help to improve the validity of the model.In the case of sudden changes in stock index monitoring,this research system shows excellent spatial structure traceability ability of fast positioning and efficient cause finding.