Memory in financial time series: From investor behavior to artificial neural networks'

M. TERESA SORROSAL-FORRADELLAS; M. GLÒRIA BARBERÀ-MARINÉ; LISANA B. MARTINEZ; MARÍA-JOSÉ GARBAJOSA-CABELLO

Reus

Congreso; XVII SIGEF Congress: Methods for decision making in an uncertain environment; 2012

Universitat Rovira i Virgili

Related to the concept of memory, the value of an asset in the past and the learning experience are key elements which influence investor behavior and, consequently, the future evolution of financial prices. In this study we focus on memory of time series, that is to say, in the way that past observations affect the future values of a financial variable. We use artificial neural networks because of its flexibility to adjust to any type of functional relationship, without the need to define it previously. As a specific case, we apply a backpropagation network to predict the price of the IBEX-35 index. Our results suggest that future prices depend on past prices, but this influence is not permanent.