A2042
Title: Long-term prediction of the metals prices using non-Gaussian time-inhomogeneous stochastic process
Authors: Agnieszka Wylomanska - Wroclaw University of Science and Technology (Poland)
Lukasz Bielak - Wroclaw University of Technology (Poland) [presenting]
Abstract: Stochastic models traditionally used to describe metals prices have proved not to be suitable to represent the dynamic behaviour and time-related nature of metal markets. Rates of return are characterized by non-Gaussian and heterogeneous characteristics, which requires the use of properly adjusted models. We introduce a stochastic model that takes under consideration mentioned specific characteristics of the real data corresponding to the mineral commodity prices, namely the non-homogeneous character (time-dependent characteristics) and non-Gaussian distribution. The introduced model is in some sense the extension of the classical Ornstein-Uhlenbeck process (called also the Vasicek model) which was originally used to the interest rate data description. The proposed model, in contrast to the classical process, has the time-dependent parameters. This perfectly captures the time-dependent characteristics of the real data. Moreover, it is based on the general class of the skewed Students t-distribution (SGT), which is related to the non-Gaussian behaviour of the real metals prices. We demonstrate here the step-by-step procedure of the time-dependent parameters estimation and check its effectiveness by using the simulated data. Finally, based on the real-time series analysis, we demonstrate that the proposed stochastic model is universal and can be applied to metals prices description for the long-term prediction.
