Dresden 2009 – scientific programme

Parts | Days | Selection | Search | Downloads | Help

AGSOE: Arbeitsgruppe Physik sozio-ökonomischer Systeme

AGSOE 14: Poster Session

AGSOE 14.4: Poster

Wednesday, March 25, 2009, 18:10–20:00, P1B

Time series processing via independent component analysis and financial asset allocation — •Sergio Rojas — Physics Department, Universidad Simón Bolívar, Valle de Sartenejas, Edo. Miranda, Venezuela

A fundamental problem in time series analysis is to find suitable representation of the signals in terms of basis that could help in extracting useful information from the data and/or to provide a better appropriate representation of the observed signals for further analysis. Linear methods widely used for this purpose include the Fourier, Haar, and cosine transformations. In this work we will examine the implementation of the relatively new technique known as Independent Component Analysis, which is intended to find non gaussian statistically independent representations of time series. By means of synthetic data that reflect some of the structural features of financial time series (like stock prices) we will show the robustness and appropriateness of the aforementioned technique for analyzing noise, incomplete and irregularly sampled time series. After that, we will address the suitability of the technique to building diversified investment financial portfolios and its applications to risk management tasks.