Research On The Measurement And Testing Of Data Stream Correlation Based On MIC

In the background of big data,the correlation mining between variables is becoming more and more important,and at the same time,it is possible to mine the data stream due to the storage capacity improvement of computer.However,the correlation analyze method of data stream is not be generalizable for the measure of nonlinear correlation,and seriously affects the measure accuracy;the correlation test method is complex,and difficult to identify the nonlinear correlation between data stream quickly and efficiently.Therefore,in order to find a correlation analysis method of nonlinear correlation of convective data,taking into account the real-time requirements of data stream mining becomes an urgent problem to be solved in the current data flow correlation measure.The research contents are as follow: firstly,this paper gives the concept of data stream,and takes the maximum coefficient of information compared with other methods,studies the applicability of MIC in measuring and testing the correlation of data stream on the basis of the characteristics of the maximum information coefficient;secondly,introduce the method system of measuring and testing of nonlinear correlation based on the maximum information coefficient,the nonlinear correlation testing method of data stream on the maximum information coefficient can accurately measure the nonlinear correlation of data stream;and compared with other testing methods,the nonlinear correlation testing method of data stream on the maximum information coefficient have the value of the test statistic and simple calculation characteristics,and can be carried out quickly and efficiently on the nonlinear correlation testing;Thirdly,through the simulation test,the effectiveness of the nonlinear correlation measuring and testing method of data stream on the MIC.Fourthly,do the empirical analysis on the relationship between HS300 index,I1 oo index and JC100 index.The main conclusions of this paper are including: firstly,using the maximum information coefficient method can solve the limitation that the traditional correlation coefficient cannot measure the nonlinear correlation,and prove that the time series data stream is also applicable;secondly,according to the equivalence characteristic of the maximum information coefficient,is used to test the nonlinear correlation of time series as design statistic,which overcome the complexity of the traditional modeling method to test the nonlinear correlation,and prove that the method is suitable for dynamic test of nonlinear correlation of time sequence data streams;thirdly,experimental studies further show that the method has a remarkable effect in the application of the dynamic relationship between stock market data.