'Northeast Volatility Wind' Effect Evolution Research by Using Neural Networks
19th International Conference Mathematical Modelling and Analysis: Abstracts of MMA2014 2014
Andrejs Pučkovs, Andrejs Matvejevs

The ’North-East Volatility Wind’ effect, desribed in previous publications is selected for subsequent research. ’North-East Volatility Wind’ effect is described as effect of volatility transmission from low-frequncy components of the signal to igh-frequency components. According results published before in financial time series a ’North-East Volatility Wind’ effect is observed. This effect is traceable by using special approach approach applicable for stock indexes, allowing to reveal the instability of financial time series initially. This approach is based on time series (signal) decomposition into components by using wavelet filtering with subsequent volatility evolution research of each signal component. According to research, a slight increase in volatility in the low-frequency components of the signal leads to significant disturbances in high-frequency components destine entire signal volatility growth. This approach is based on signal decomposition by using the wavelet filtering. Wavelet filtering is applied by using Direct and Inverse CWT for each scaling parameter.~\cite{3book} Thus for each scaling parameter the signal component (which is part of the original signal) is obtained. For subsequent research volatility indicator is analysed by using 20-days time window, which is shifted on the time axis. Volatility analysis is done for each signal component. As a result volatility evolution in time is obtained for each signal component. As a next step volatility evolution crosscorrelation analysis is done for each signal component in order to describe volatility transmission from one volatility component to another. In other words volatility transmission from one volatility layer to another is analysed. In current research volatility evolution crosscorrelation analysis is extended by using Neural Network Algorithms, to trace volatility evolution dependences in time, optimising Neural Network structure and finding correspondent weights. In current research complicated relationships between volatility layer are discovered. As a result ’North-East Volatility Wind’ Effect brings out deeper understanding of volatility evolution and opportunity to illuminate most dramatical market drawdowns initially. This opportunity is explained by ability to see a very small changes in volatility logarithm in the low-frequency components of the signal.


Atslēgas vārdi
Wavalet analysis, Nueral Networks, Continiouse Wavelet Transform, Time Series, Volatility, Time Series Analysis

Pučkovs, A., Matvejevs, A. 'Northeast Volatility Wind' Effect Evolution Research by Using Neural Networks. No: 19th International Conference Mathematical Modelling and Analysis: Abstracts of MMA2014, Lietuva, Druskinkai, 26.-29. maijs, 2014. Druskininkai: Vilnius Gediminas Technical University (VGTU), 2014, 54.-54.lpp. ISBN 9786094576928.

Publikācijas valoda
English (en)
RTU Zinātniskā bibliotēka.
E-pasts: uzzinas@rtu.lv; Tālr: +371 28399196