Multidimensional scaling analysis of financial time series based on modified cross-sample entropy methods

摘要: Abstract Stocks, as the concrete manifestation of financial time series with plenty potential information, are often used in study series. In this paper, we utilize stock data to recognize their patterns through out dissimilarity matrix based on modified cross-sample entropy, then three-dimensional perceptual maps results provided multidimensional scaling method. Two methods proposed that is, Kronecker-delta entropy (MDS-KCSE) and permutation (MDS-PCSE). These two use replace distance or measurement classical (MDS). Multidimensional Chebyshev (MDSC) is employed provide a reference for comparisons. Our analysis reveals clear clustering both synthetic 18 indices from diverse markets. It implies generated by same model easier have similar irregularity than others, difference index, which caused country region different policies, can reflect data. experiments, not only models be distinguished, one under parameters also detected. experiment, clearly divided into five groups. Through analysis, find they correspond regions, respectively, Europe, North America, South Asian-Pacific (with exception mainland China), China Russia. The demonstrate MDS-KCSE MDS-PCSE more effective divisions experiments MDSC.