Research On The Identification Of Stock Market Change Point Structure Under Concept Drift Mapping

The problem of change point identification has been widely used in the financial field since the introduction of Page,especially for describing and analyzing the phenomenon drift phenomenon in the stock market.By identifying the structural changes in the stock market's rate of return,volatility and volume,and analyzing the causes of the changes,we can more accurately grasp the operating state of the stock market and improve the risk management ability of the stock market.Especially during the period before the 70 th anniversary of the founding of the financial crisis,the European debt crisis,the listing of the GEM,the launch of margin financing and stock index futures,the opening of Shanghai-Hong Kong Stock Connect and Shenzhen-Hong Kong Stock Connect,and the Sino-US trade war It has brought certain influence to the economy,causing the stock market to have ups and downs and fluctuations,and the concept drift phenomenon occurs,thus causing structural changes in the stock market's yield,volatility and volume series.In view of the above background,when introducing the gains,fluctuations and trading volume of the Chinese and American stock markets,the flow identification method-concept drift map is introduced to accurately identify the structural change points of the two stock markets,and the change point analysis is made,and then the Chinese and American stock markets are The structural characteristics of the variable are compared.Firstly,this paper adopts the CSI 300 Index and the S&P 500 Index from October 1,2009 to September 30,2019,every five minutes of high frequency data,using the concept drift mapping model,utilizing the stream identification technology to accurately identify the changes point in the stock market's yield,volatility and volume,and analyze the causes of the change point.Secondly,the paper compares and analyzes the difference of the change point structural characteristics between Chinese and American stock market.Finally,the variable structure Copula model is used to analyze the linkage of change point structure of the two stock markets from the static and dynamic perspectives.The following conclusions are drawn: firstly,the concept drift mapping model is applied to identify4,2 and 1 change points of yield,volatility and volume in Chinese stock market respectively,and the time of structural mutation can correspond to major economic events.Secondly,the difference and linkage exist between the variable structure characteristics of Chinese and American stock markets.From the perspective of difference,the number of change points in the yield and volatility of the Chinese stock market are more than that of the US stock market,with 3,1 more points respectively,indicating that the Chinese stock market has changed more frequently.From the perspective of linkage,the Copula correlation coefficient Theta between Chinese and the US stock market is 1.595,and the Copula correlation coefficient Theta of the change point structure is 1.392 and 1.798 respectively,indicating that there is a certain linkage between the change point structures of the two stock markets.The innovations of this paper are as follows: on the one hand,the stream identification technology is integrated with the concept drift-mapping model,applied to the stock market,quantifying the drift phenomenon of the stock market,calculating the drift magnitude,and identifying the point where the drift magnitude is equal to 1 as the change point.The model determines whether the change point is a change point by measuring the drift magnitude,instead of determining the existence of the change point by the test method,so that the recognition result is more effective and accurate.On the other hand,this paper uses the 5-minute high-frequency data of the China-US stock index,which has the characteristics of real-time streaming data.Compared with the low-frequency data,the discrete model based on low-frequency data will cause information loss.Based on this,the strategy model and trend analysis will lack accuracy and affect investment judgment.High-frequency data contains useful information about the market's microstructure,and it can analyze the conceptual drift of the stock market from different dimensions.