Abstract:
"
Emerging markets like Sri Lanka are becoming an attractive place for foreign investors. Foreign investment
mainly depends on the forecasted exchange rate movements. This study attempts to determine, how
accurately a multiple regression model would predict the USD/LKR exchange rate considering most
influential economic factors outlined by economists. The specific objective of this study is to determine the
best model. This is done by analyzing forecasting performance of various regression models. Eight models
including the Multiple Linear Regression (MLR), Lasso Regression, Elastic Net Regression, Ridge
Regression, Support Vector Regression, Decision Tree Regression, Random Forest Regression and Vector
Auto Regression. All these forecast models are executed using monthly data over the period January 2000
to December 2020.
This study undertakes an econometric analysis of determinants of exchange rate for US Dollar in terms of
Sri Lankan Rupee within the framework of monetary approach. Worker’s Remittance, Gross Official
Reserves, Fuel Import, Machinery & Equipment Import, Tea Export, Apparel Export, Stock Turnover and
Local Debt are taken as determinants of USD/LKR exchange rate during the managed floating regime in
Sri Lanka. Exploratory Data Analysis (EDA) method in Python is used to identify the most significant
features of the dataset and to handle missing values and make transformations to the dataset as needed.
The machine learning models were trained and tested to predict the response variable USD/LKR Exchange
Rate. The Multiple Linear Regression, Ridge Regression, Elastic Net Regression, VAR and Random Forest
yielded high accuracy levels predicting the exchange rate whereas the Decision Tree, SVM Regression
model and Lasso Regression poorly performed and that could be due to the low magnitude of the dataset.
Given there is no standardized framework to predict the USD/LKR Exchange Rate in Sri Lanka, from this
study it was proven that based on the current economic situation of the country, even the highly influential
factors rather than going for broader parameters (i.e. Imports Total and Exports Total) can have a strong
predictive ability on the Exchange Rate in Sri Lanka."
