Comparison of Machine Learning Models for Value at Risk Calculations in the Chemistry Index
Keywords:
Machine Learning, Value at Risk, Istanbul Stock ExchangeAbstract
Value at Risk (VaR) calculation is one of the critical issues in portfolio management. As the calculation system becomes more complex, there is an increasing need for computer applications. In VaR calculation using machine learning has become a commonly used method. With machine learning, VaR can be calculated using various methods, different ML algorithms, and varying time horizons. This flexibility allows for more robust and adaptable risk assessment in portfolio management.
In this study, a portfolio was constructed by using the four largest-volume stocks in the chemistry index (XKMYA) of Borsa İstanbul (Istanbul Stock Exchange), namely PETKM, HEKTS, SASA, and TUPRS. The distribution of these stocks in the portfolio was determined using Monte Carlo simulation. The study utilized the Parametric VaR method to calculate risk for a 10-day, 3-period timeframe.
To achieve this, daily closing stock price data spanning 5 years from August 24, 2018, to August 24, 2023, was employed. In the comparative analysis, machine learning models, including Random Forest (RF), Support Vector Machine (SVM), Decision Trees (DT), and Linear Regression (LR), were compared. Since the comparison was based on predictions, error metrics such as RMSE, MSE, MAE, and MAPE were used to measure the efficiency of the models.
The analysis revealed that the RF model provided the best results for the prepared portfolio.
References
Abar, H. (2020). GARCH Modeli ve DVM–EKK Regresyonu ile Kripto Para Fiyat Öngörüsü: Bitcoin Fiyatı Üzerine Bir Uygulama. Turkish Studies - Economics Finance Politics, 15(2), 705-725.
Bayer, S. (2018). Combining value-at-risk forecasts using penalized quantile regressions. Econometrics and statistics, 8, 56-77.
Behera, J., Pasayat, A. K., Behera, H., & Kumar, P. (2023). Prediction based mean-value-at-risk portfolio optimization using machine learning regression algorithms for multi-national stock markets. Engineering Applications of Artificial Intelligence, 120, 105843.
Bozsik, J. (2011). Decision tree combined with neural networks for financial forecast. Periodica Polytechnica Electrical Engineering (Archives), 55(3-4), 95-101.
Duman, S., & Türkmen, S. Y. (2023). Bankacılık Sektöründe Kredi Risk Yönetimi ve Bir Tahmin Modeli Örneği. Finansal Araştırmalar ve Çalışmalar Dergisi, 15(28), 1-14.
Esi, A. M. (2022). Bankacılık Sektöründe Kredi Ödemelerinin Makine Öğrenimi Sınıflandırma Algoritmalarına Göre Analizi (Doctoral dissertation, Marmara Universitesi (Turkey)).
Karacan, S., & Kırdar, M. (2021). Hisse Senedi Fiyat Tahmininde Makine Öğrenmesi Ve Yapay Zeka Kullanımı. Journal of International Social Research, 14(76).
Kuester, K., Mittnik, S., & Paolella, M. S. (2006). Value-at-risk prediction: A comparison of alternative strategies. Journal of Financial Econometrics, 4(1), 53-89.
Kumar, M., & Thenmozhi, M. (2006, January). Forecasting stock index movement: A comparison of support vector machines and random forest. In Indian institute of capital markets 9th capital markets conference paper.
Sarı, S. S., & Başakın, E. E. (2021). Borsa İstanbul Banka Endeksi’nin Veri Tabanlı Modeller ile Analiz Edilmesi. Verimlilik Dergisi, (3), 147-163.
Sarı, S. S., & Yiğiter, Ş. Y. (2020). Yatırımcı Duyarlılığının Hisse Senedi Getirilerindeki Rolü ve Tüketici Güven Endeksiyle Ölçülmesi. Atatürk Üniversitesi İktisadi ve İdari Bilimler Dergisi, 34(1), 77-97.
Tütüncü, T. E., & Gürsakal, S. (2023). Kredi Temerrüt Riskini Tahmin Etmede Makine Öğrenme Algoritmalarının Karşılaştırılması. Avrupa Bilim ve Teknoloji Dergisi, (50), 14-22.
Westgaard, S., Århus, G. H., Frydenberg, M., & Frydenberg, S. (2019). Value-at-risk in the European energy market: a comparison of parametric, historical simulation and quantile regression value-at-risk. Journal of Risk Model Validation, 13(4), 1-27.
Zhang, H. G., Su, C. W., Song, Y., Qiu, S., Xiao, R., & Su, F. (2017). Calculating Value-at-Risk for high-dimensional time series using a nonlinear random mapping model. Economic Modelling, 67, 355-367.
