Classification Techniques applied to the study of high frequency data arising in finance and geophysics

Maria Pia Beccar-Varela

  The University of Texas at El Paso

Abstract:

In this work, we apply four classification techniques: Dynamic Fourier Transform, Wavelet Transform, Discriminant Analysis based on Chernoff and Kullback-Leible distances that measure the similarity between statistical populations, and Clustering algorithms to identify the signals of occurrences arising in financial crashes, and in earthquakes and explosive data.

Specifically, we analyzed the sequence of intraplate earthquakes with 5.2 magnitude that occurred in Arizona in 2014 at the same location where mining explosions was performed some years earlier. We also studied the high frequency data (HFD) corresponding to the collapse of two market crashes: the Lehman Brothers collapse on September 15, 2008 and the Flash Crash event on May 6, 2010. The application of these methodologies to the Lehman Brothers collapse and the Flash Crash event indicate that the Lehman Brothers collapse event behaves like a natural tectonic earthquake while the Flash Crash event behaves like a human made explosion.

This conclusion suggests that since the Lehman Brothers event behaves like a natural tectonic earthquake, it is predictable, whereas the Flash Crash event is not predictable since it behaves like a human made explosion. Our analysis is consistent since the Flash Crash event was due to human error hence not predictable.

We conclude that the four methodologies can effectively distinguish and identify a seismic event and a human made explosion. Furthermore, the results are consistent with the results previously found when applying only wavelet techniques to the two set of data.

This presentation reflects joint work with I. Florescu, M.C. Mariani, M.A.M Bhuiyan and O. Tweneboah.

References:

Maria P. Beccar-Varela, Maria C. Mariani, Osei K. Tweneboah and Ionut Florescu, “Analysis of the Lehman Brothers collapse and the Flash Crash event by applying wavelets methodologies”, Physica A: Statistical Mechanics and its Applications, Volume 474, pp 162-171, 2017.

Maria P. Beccar-Varela, Hector Gonzalez-Huizar, Maria C. Mariani, and Osei K. Tweneboah “Use of wavelets techniques to discriminate between explosions and natural earthquakes”, Physica A: Statistical Mechanics and its Applications, 457, pp 42-51, 2016.