A Study on the Influence of the Discretisation Unit on the Effectiveness of Modelling Currency Exchange Rates Using the Binary-Temporal Representation

Abstract

An exchange rate can be expressed in the form of a binary-temporal representation. Such a representation is based on a discretization of movements in the exchange rate, in which to each change in the value - equal to a given discretization unit - two parameters are allocated: a binary value, consistent with the direction of change in the exchange rate (increase 1, decrease 0) and duration. Statistical examination proves the existence of dependencies between the parameters of previous changes and the direction of future changes. To model the exchange rate using the applied binary-temporal representation, an appropriate model was developed that enables estimation of the probability of the direction of future changes in the currency exchange rate based on the parameters of historical changes. This article presents an analysis of the influence of the chosen discretization unit on the quality of exchange rate modelling. For this purpose, software was written in MQL4 and C++. As a result of the study, an optimal value for the discretization unit and the optimal parameters of the model providing the highest efficiency were determined. The input data used in the analysis involved tick data for the AUD/NZD exchange rate for a five-year time frame 2012-2017. (original abstract)