摘要
应用扩散熵方法分析道琼斯工业平均指数概率分布函数P(x,t)的标度行为,得到反映其内在特性的标度值,同时给出了不同时间尺度t下扩散过程的对数概率分布函数log10(P(x,t)),该分布函数中心符合Lévy分布,两端具有胖尾特征.由于扩散过程的概率分布函数是非Gaussian型,因此应用扩散熵方法分析其标度行为优于其他方法,而采用实数阶矩估计的标度值明显偏离扩散熵所计算的值.
The concept of diffusion entropy (DE) was used to estimate the scaling behavior. With this method, the scaling value of Dow Jones industrial average index was obtained. Then, the log-probability distribution function log10 (P(x,t)) of the diffusion process was obtained by time series with different t. It obeys the centered Levy distribution and displays fat-tail property, which means that the estimating methods on the basis of moments are unavailable. However, the scaling value given by the method of fractional derivative analysis (FDA) appreciably deviates from the genuine value given by diffusion entropy.
出处
《中国科学技术大学学报》
CAS
CSCD
北大核心
2006年第12期1281-1284,1320,共5页
JUSTC
基金
国家自然科学基金(70271070
70471033
10472116
10532060
10547004
70571074
70571075)资助
关键词
金融复杂性
标度值
扩散熵
实数阶矩分析
financial complexity
scaling value
diffusion entropy
fractional derivative analysis