金融数据挖掘中的非线性相关跟踪技术_英文_.pdfVIP

1000-9825/ 2000/ 11( 12) 1581-06 c2000 Journal of Softw are Vol. 11, No . 12 Nonlinear Correlation Tracking Technique in Data Mining of Financial Markets - , - , - Y I Don yun ZHANG Wei min DU Xiao yon ( Departm ent of Mathematics and Sy stem Science, Nat ional University of Defense T echnolo y, Chan sha 410073, China) - : @ . . E mail dyyi nudt edu cn Received November 25, 1998; accept ed October 18, 1999 Abstract: Financial data minin is one of the most challen in research directions in information societ y. Financial data w ith random characteristics m ake it difficult t o find out the rule hidden in data . In this paper , it , is pointed out t hat correlation coefficient can not capture nonlinear information which is the serious defect of classic correlation analysis. Furtherm ore, the properties of the hi h-order correlation coefficient are dis- cussed, and it is proved that hi h-order correlation can not only describe the hidden nonlinear correlation, but also fill up t he space betw een classic correlation and independence. T he computational simplicity makes the hi h-order correlation coefficient be an effective technique to track nonlinear relation betw een variables. Final- ly , t he above result s are applied t o the correlative analy sis betw een stock price and stock tradin volume, and the computin results show that the hi h-order correlation coefficient can track the time-vary in nonlinear characteristics. Key words: nonlinear analysis; data minin ; financial