Summary

International Symposium on Nonlinear Theory and Its Applications

2016

Session Number:B3L-C

Session:

Number:B3L-C-5

Forecasting Correlation Structures

Martin Schule,  Thomas Ott,  Peter Schwendner,  

pp.-

Publication Date:2016/11/27

Online ISSN:2188-5079

DOI:10.34385/proc.48.B3L-C-5

PDF download (160.4KB)

Summary:
Often the signature of a complex system is a number of empirically found time series. As the exact processes generating these series are often unknown, a key tool to investigate the structural behaviour of the complex system is considering the correlation structure, i.e. the system of estimated pairwise correlations between the time series. As the correlation coefficients are calculated from the given data set, there is a priori no way to forecast the future behaviour of the system. However, if there are certain consistent patterns in the analyzed correlation structure, one may estimate the future correlations up to a certain point. The contribution presents a general method, based on eigenmode oscillations analysis and multivariate time series forecasting techniques, allowing to forecast correlation structures.