Undoubtedly, the COVID-19 pandemic has caused the economic landscape around the world to be in an unprecedented state of continuous and heightened fluctuation in recent years and to this very day. This economic shift has prompted countries around the world including Malaysia to recentre their economic recovery attention to the dynamic currency exchange rate movement which is greatly associated with economic growth. For instance, recent news reported the worrying depreciation of the exchange rate of MYR and JPY against USD to a near-record low in the economic recovery period due to external economy-pressuring circumstances, with MYR 1 being equivalent to USD 0.21 as of 15 April 2024. This can have a detrimental impact on the economic competitiveness of Malaysia on the global stage as well as foreign investment and economic growth.
Therefore, this calls for the paramount need to predict currency exchange rates of MYR more reliably and accurately based on economic indicators such as GDP, unemployment and interest rate as well as prices of commodities including rubber and palm oil using time-series statistical analysis as well as machine and deep learning techniques so that the worsening currency depreciation in the post-pandemic era which is filled with complexities can be effectively mitigated. In terms of implications, various stakeholders can benefit from the importance of forecasting the MYR exchange rate against other currencies based on economic indicators. The decision-making process of policymakers, investors and other business stakeholders can be facilitated based on the exchange rate prediction, thus implementing more effective national financial policies and investment risk management plans that ultimately benefit the economic environment of Malaysia.
The problem to be addressed is the exclusion of different economic attributes such as GDP, unemployment and interest rate as well as prices of commodities including rubber and palm oil in predicting the currency exchange rate. Multiple past studies have focused on using different time-series statistical models and machine and deep learning methods for univariate accurate prediction of the exchange rate of one currency against another based on their open and close prices only. Nevertheless, the impact of economic indicators on the movements of currency exchange rates in the currency market can affect the accuracy of the prediction results, which most studies did not consider. This can result in ineffective monetary policies and investment risk management strategies which may cause the economy of Malaysia to go downhill. Besides that, the unprecedented heightening of current exchange rate movement fluctuations due to the COVID-19 pandemic indicates that the data patterns are more complex and filled with more non-linearity and non-stationary, which these degrees can be heightened even more with the inclusion of economic variables. Therefore, one proposed solution in this project is the use of machine and deep-learning techniques to compare and accurately forecast the MYR exchange rate against other currencies with the economic indicators as the model inputs.
The objectives of this project are as follows:
-
To conduct the development of baseline models using techniques such as SVR, MLP, CNN and LSTM and identify the baseline models with the best performance in forecasting the exchange rate of MYR against USD, EUR, GBP, SGD and CNY based on MAE, MSE, RMSE and MAPE
-
To conduct the development of a hybrid model by stacking LSTM with CNN to identify the hybrid models with the best performance in forecasting the exchange rate of MYR against USD, EUR, GBP, SGD and CNY based on MAE, MAE, MSE, RMSE and MAPE
-
To compare the performance between the most optimal baseline and hybrid model based on MAE, MSE, RMSE and MAPE
-
To discover whether interest, unemployment and inflation rates as well as gold, rubber, crude and palm oil prices are the most important economic factors in forecasting the exchange rate of MYR against USD, EUR, GBP, SGD and CNY