CDS Spreads Forecast: An Application on Türkiye

Abstract

Purpose – Development of a Model for Predicting Daily CDS Premiums: A Study on the Turkish Financial Market. This research aimed to develop a model for predicting daily Credit Default Swap (CDS) premiums. The primary goal of this study is to contribute to the prediction of CDS premiums, which serve as an indicator when making investment decisions. In the context of Turkey, where financial indicators exhibit high volatility, CDS premiums stand out as a crucial metric reflecting the perception of risk in financial markets and serving as an indicator for investment decisions, contributing significantly to the determination of economic stability. Design/Methodology/Approach - In this study, artificial neural networks, a method within the discipline of machine learning, have been utilized. Three prediction models have been developed using artificial neural networks. Findings – Although all models in the study exhibited successful prediction performance with low mean absolute error (MAE) rates, Model 2 demonstrated the best performance. Additionally, the machine memorization problem was tested using a confusion matrix. Discussion – In countries with high volatility, an alternative method has been developed for predicting CDS premiums. While traditional financial models often struggle to provide stable forecasts in such markets, the proposed method ensures a more accurate consideration of uncertainty and fluctuations. At the core of this alternative approach lies the operation of an algorithm capable of learning volatility and risk movements.