Show/Hide Menu
Hide/Show Apps
anonymousUser
Logout
Türkçe
Türkçe
Search
Search
Login
Login
OpenMETU
OpenMETU
About
About
Open Science Policy
Open Science Policy
Frequently Asked Questions
Frequently Asked Questions
Communities & Collections
Communities & Collections
Modeling of exchange rates by multivariate adaptive regression splines and comparison with classical statistical methods
Download
index.pdf
Date
2017
Author
Köksal, Ece
Metadata
Show full item record
Item Usage Stats
6
views
3
downloads
Economic factors like inflation, interest rates and exchange rates are among the leading indicators of a country’s relative level of economic health. With the help of technological improvements and global requirements, trading volume and a wide range of commerce network, exchange rates play a vital role in economics and finance since a higher exchange rate may result in a lower trade balance of a country, whereas a lower rate may cause an increase. Inflation, interest rates, domestic money supply growth, a country’s balance of payments’ size and trend, a country’s economic growth, dependency on outside sources and central bank intervention, are the factors which affect an exchange rate. Since many dependent and independent factors affect exchange rates, it is difficult to predict them. In areas of application, data mining is frequently used for decision support, financial forecasting, marketing strategy, prediction, etc. The method of data mining and machine learning is applied to analyze and forecast the future behavior of such complex systems. Modeling and prediction of exchange rates are still a challenge, although mathematicians, economists and statisticians have worked to reach a model with a superior forecasting ability for many years. Therefore, in this study, we aim to generate mathematical models to forecast the monthly US Dollar (USD) / Turkish Lira (TRY) and Euro (EUR) / Turkish Lira (TRY) exchange rates via data mining tools. For this purpose, we apply a flexible model Multivariate Adaptive Regression Splines (MARS) and widely used models Linear Regression (LR) and Support Vector Regression (SVR). In this study, MARS, LR and SVR models applied on USD / TRY and EUR / TRY exchange rate data sets in the period of 01/01/2007 and 30/04/2015; then the results of these models are compared and found out that MARS method has superior forecasting ability over LR and SVR methods for USD / TRY and EUR / TRY exchange rates. The thesis ends with a conclusion and an outlook to future investigations.
Subject Keywords
Foreign exchange.
,
Regression analysis .
,
Mathematical statistics.
,
Foreign exchange rates.
,
Foreign exchange market.
URI
http://etd.lib.metu.edu.tr/upload/12621053/index.pdf
https://hdl.handle.net/11511/26705
Collections
Graduate School of Applied Mathematics, Thesis