Intraday analysis of portfolios on the ASX using ICA

Abstract

In this paper we address the problem of forecasting non-gaussian portfolio returns over multiple time scales. We apply a relatively new technique for estimating portfolio returns by considering higher order mutual information. This technique is based on two methodologies: Independent Component Analysis and Gaussian mixtures. We apply this model to intraday data from the ASX. Our findings illustrate that this model is particularly useful in estimating portfolio returns over a short time scale when the distribution is highly non-Gaussian. © Springer-Verlag 2003.