Democritus University of Thrace (DUTH)
Πρόβλεψη μακροοικονομικών μεταβλητών με τη βοήθεια της καμπύλης αποδόσεων και μεθόδους μηχανικής μάθησης
Abstract
dc:descriptionSeveral studies highlight the forecasting ability of the yield curve in terms of economic activity. These studies use the information provided by the slope of the yield curve i.e. pairs of short and long term interest rates, in an effort to forecast the business cycle. In this thesis we focus our analysis in forecasting deviations of real U.S. GDP and Eurocoin index from their long-run trend paying particular attention to the instances of an output gap (recession). In doing so, we are not limiting our empirical analysis to the slope of the yield curve and construct three alternative models: one that uses only pairs of interest rates as predictors of the business cycle, and two models where we try to exploit the curvature and not only the slope of the yield curve. In these latter setups we use combinations of three or more interest rates in an effort to trace the curvature of the yield curve. In these models we use three alternative forecasting methodologies. The commonly used in this line of literature probit model and the Support Vector Machines (SVM) and neural networks approach from the area of Machine Learning. The data referring to the economy of the U.S. are quarterly and cover the period from 1976Q3 to 2011Q4 and the data of the Eurozone cover the period from 2004:9-2014:10. Our results show that we can achieve a 95.65% out-of-sample forecasting accuracy for output gaps in the case of the SVM methodology.
Degree
thesis:*- Grantor dc:publisher
- Democritus University of Thrace (DUTH)
- Year dc:date
- 2015
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Chrysanthidou, Efthimia
- Χρυσανθίδου, Ευθυμία
Subjects
dc:subject × 11Rights
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/36436