Nonlinear difference equation forecasting engine for FOREX markets.

• Main Author: Tan, James Peng Lung.
• Other Authors: Cheong Siew Ann
• Format: Theses and Dissertations
• Language: English
• Published: 2012
• Subjects: DRNTU::Business::Finance::Foreign exchange; DRNTU::Science::Mathematics::Applied mathematics::Complex systems
• Online Access: http://hdl.handle.net/10356/49505
• Institution: Nanyang Technological University

In this study, we aim to find stable models that can be used to model the exchange rate time series using non-linear difference equations. Because the FOREX market is a dynamical system, we can expect the exchange rate to switch from model to model. Therefore, we can develop a forecasting engine that can switch between the best fitting models to obtain reliable forecasts of the exchange rate. To do this, we first overfitted the model to capture as many relevant parameters as possible and then through sensitivity analysis, filtered out the least relevant parameters. Then we tracked the relevant parameters through time to determine their stable values for use in a forecasting engine. However, the parameters were found to vary too wildly and did not remain stable at all. We then found that a change in interest rate produced a deviation from the normal behavior of the model. Forecasts using a simpler non-linear difference equation with the more sensitive parameters were then carried out.