Variations in financial time series: Modelling through wavelets and genetic programming

Abstract

We analyze the variations in S&P CNX NSE daily closing index stock values through discrete wavelets. Transients and random high frequency components are effectively isolated from the time series. Subsequently, small scale variations as captured by Daubechies level 3 and 4 wavelet coefficients and modelled by genetic programming. We have smoothened the variations using Spline interpolation method, after which it is found that genetic programming captures the dynamical variations quite well through Padẽ type of map equations. The low-pass coefficients representing the smooth part of the data has also been modelled. We further study the nature of the temporal variations in the returns.