Energy CDS Premium Prediction: Evidence from Comparative Study of Machine Learning Algorithms
19th IEEE International Conference on Application of Information and Communication Technologies, AICT 2025, Al Ain, Birleşik Arap Emirlikleri, 29 - 31 Ekim 2025, (Tam Metin Bildiri)
- Yayın Türü: Bildiri / Tam Metin Bildiri
- Doi Numarası: 10.1109/aict67988.2025.11268757
- Basıldığı Şehir: Al Ain
- Basıldığı Ülke: Birleşik Arap Emirlikleri
- Anahtar Kelimeler: ANN, Decision Tree, Energy CDS, GAM, Machine Learning, Random Forest, SVM, XGBoost
- Ankara Yıldırım Beyazıt Üniversitesi Adresli: Hayır
Özet
In the energy sector, credit default swap (CDS) premiums are widely regarded as a critical indicator of financial risk, playing an essential role in guiding investor decision-making. This study aims to improve the predictability of energy CDS premiums by comparing various machine learning algorithms. A daily dataset of 5,502 observations was used, covering the period from 01.01.2004 to 31.01.2025. The dataset was divided into a training set (01.01.2004–31.12.2023) and a validation set (01.01.2024–31.01.2025, comprising 267 observations). To better capture the persistence of CDS dynamics and short-term market reactions, a five-lag structure was incorporated into all models. The predictive performance of Artificial Neural Networks (ANN), Support Vector Machines (SVM), XGBoost, Generalised Additive Models (GAM), Random Forests (RF) and Decision Trees (DT) were compared. The results reveal that all methods achieved a high level of accuracy, with mean absolute percentage error (MAPE) values consistently below 2%, indicating that CDS premiums can be reliably forecasted using these techniques. Nevertheless, GAM outperformed all other approaches by delivering the lowest MAPE, MSE and MAE values, as well as the highest R2. These findings highlight the advantage of methods capable of modelling nonlinear relationships flexibly in the forecasting of CDS premiums, providing both methodological contributions to the literature and practical implications for risk management and strategic decision-making in energy markets.