Calculating Hurst exponent and neutron monitor data in a single parallel algorithm

Abstract

We implemented an algorithm for simultaneous parallel calculation of the Hurst exponent H and the fractal dimension D for the time series of interest. Parallel programming environment was provided by OpenMPI library installed on three machines networked in the virtual cluster and operated by Debian Wheeze operating system. We applied our program for a comparative analysis of week and a half long, one minute resolution, six channels data from neutron monitor. To ensure a faultless functioning of the written code we applied it to analysis of the random Gaussian noise signal and time series with manually introduced self-affinity features. Both of them have the well-known values of H and D. All results are in good correspondence with each other and supported by the modern theories on signal processing thus confirming the validity of the implemented algorithms. Our code could be used as a standalone tool for the different time series data analysis as well as for the further work on development and optimization of the parallel algorithms for the time series parameters calculations.