| Abstract |
Much of the literature on financial forecasting assumes that the return distribution is log normal. However, in the case of some commodities (e.g. crude oil) it is not so. In the case of crude oil it has been observed that the log of the returns distribution is both skewed and has high kurtosis (leptokurtic). Even after accounting for structural breaks data show very high kurtosis. In such situations standard forecasting methodologies based on normal distribution cannot be used. In this project we will attempt to compare three new methodologies for price forecasting with highly leptokurtic data. |