SUKUK ENDEKSLERİNDE DERİN ÖĞRENME İLE GETİRİ TAHMİNİ: BİLSTM, GRU VE RNN MODELLERİNİN KARŞILAŞTIRMASI

Abstract

Bu çalışma, finansal zaman serilerinin ileriye dönük tahmininde üç farklı derin öğrenme mimarisinin (BİLSTM, GRU, RNN) performanslarını karşılaştırmalı olarak incelemektedir. 12.02.2014–04.07.2025 dönemine ait Bloomberg, DJSUKUK ve SPBDMSUT endekslerinin günlük getiri serilerinden oluşan veri seti üzerinde yapılan analizde, her model 60 günlük geçmiş verilere dayanarak 20 günlük ileri tahmin üretmiştir. Modeller, MAE, RMSE ve R² gibi regresyon tabanlı metriklerle değerlendirilmiş ve tahmin çıktıları grafiksel olarak analiz edilmiştir. Ön analizler kapsamında durağanlık testleri (ADF), otokorelasyon analizleri (ACF/PACF), korelasyon matrisleri ve tanımlayıcı istatistikler kullanılmıştır. Bulgular, serilerin güçlü kısa vadeli bağımlılıklara ve birlikte hareket eden örüntülere sahip olduğunu ortaya koymuştur. BİLSTM modeli, tüm değişkenlerde en düşük hata ve en yüksek determinasyon katsayısını (R²) üretmiş, özellikle SPBDMSUT ve DJSUKUK serilerinde güçlü performans göstermiştir. GRU modeli bazı durumlarda rekabetçi sonuçlar üretmiş, RNN ise belirgin şekilde en düşük performansı sergilemiştir. Sonuçlar, finansal tahmin problemlerinde yapay zekâ tabanlı modellerin yüksek doğruluk sağladığını, özellikle BİLSTM mimarisinin karmaşık ve çok değişkenli serilerde öne çıktığını göstermektedir. Elde edilen bulgular, finansal karar destek sistemleri, politika planlama ve risk yönetimi alanlarında derin öğrenme tabanlı yapıların stratejik avantajlar sunduğunu ortaya koymaktadır.

Keywords

• BILSTM
• GRU
• RNN
• Finansal Tahmin
• Sukuk Piyasası.

RETURN FORECASTING IN SUKUK INDICES USING DEEP LEARNING: A COMPARISON OF BILSTM, GRU, AND RNN MODELS

Abstract

This study presents a comparative evaluation of the forecasting performance of three deep learning architectures: BİLSTM (Bidirectional Long Short-Term Memory), GRU (Gated Recurrent Unit), and RNN (Recurrent Neural Network), applied to financial time series. The dataset includes daily return series from Bloomberg, DJSUKUK, and SPBDMSUT indices, covering the period between 12 February 2014, and 4 July 2025. Each model was trained on 60 days of historical data to forecast the following 20 days. Model performance was assessed using regression-based metrics, including MAE, RMSE, and the coefficient of determination (R²), while forecast outputs were also examined graphically. Initial steps involved unit root testing (ADF), autocorrelation checks (ACF/PACF), correlation matrices, and descriptive statistics. The analyses revealed strong short-term dependencies and interconnected patterns across the variables. Among the tested models, BİLSTM consistently achieved the lowest prediction errors and highest R² values, particularly excelling for the SPBDMSUT and DJSUKUK indices. GRU performed competitively in some cases but generally fell short of BİLSTM, while RNN showed the weakest predictive accuracy overall. These results highlight the practical usefulness of deep learning models in financial forecasting, positioning BİLSTM as a particularly effective tool for handling complex and multivariate time series. The findings suggest that such models can offer valuable support for financial decision-making, policy design, and risk management activities.

Keywords

• BILSTM
• GRU
• RNN
• Financial Forecasting
• Sukuk Market

References

1. Abdel-Khaleq, A. H., & Richardson, C. F. (2007). New horizons for Islamic securities: Emerging trends in sukuk offerings. Chicago Journal of International Law, 7(2), 409–425. https://chicagounbound.uchicago.edu/cjil/vol7/iss2/5/
2. Arora, A., Bajpai, S., & Prakash, M. (2023). Cryptocurrency price prediction using LSTM neural networks. In International Conference on Recent Trends in Data Science and Its Applications (pp. 679–682). River Publishers. https://doi.org/10.13052/rp-9788770040723.130
3. Arundina, T., Kartiwi, M., & Omar, M. A. (2016). Artificial Intelligence for Islamic Sukuk Rating Predictions. 211-241. https://doi.org/10.1057/978-1-137-48880-0_8
4. Bengio, Y. (2009). Learning deep architectures for AI. Foundations and Trends in Machine Learning, 2(1), 1–127. https://doi.org/10.1561/2200000006
5. Cho, K., Van Merriënboer, B., Gulcehre, C., Bahdanau, D., Bougares, F., Schwenk, H., & Bengio, Y. (2014). Learning phrase representations using RNN encoder-decoder for statistical machine translation. arXiv. https://arxiv.org/abs/1406.1078
6. Chung, J., Gulcehre, C., Cho, K., & Bengio, Y. (2014). Empirical evaluation of gated recurrent neural networks on sequence modeling. arXiv. https://arxiv.org/abs/1412.3555
7. Çetin, D. T., & Metlek, S. (2021). Forecasting of Turkish sovereign sukuk prices using artificial neural network model. Acta Infologica, 5(2), 241–254. https://doi.org/10.26650/acin.907990
8. Guresen, E., Kayakutlu, G., & Daim, T. U. (2011). Using artificial neural network models in stock market index prediction. Expert Systems with Applications, 38(8), 10389–10397. https://doi.org/10.1016/j.eswa.2011.02.068

9. Fu, Y., Lou, F., Meng, F., Tian, Z., Zhang, H., & Jiang, F. (2018). An intelligent network attack detection method based on RNN. In 2018 IEEE Third International Conference on Data Science in Cyberspace (DSC) (pp. 483–489). IEEE.
10. Han, C., & Fu, X. (2023). Challenge and opportunity: Deep learning-based stock price prediction by using bi-directional LSTM model. Frontiers in Business, Economics and Management, 8(2), 51–54. https://doi.org/10.54097/fbem.v8i2.6616
11. Hasan, M. R., Sultana, F., Rashid, H. O., Abdulwasea, T. K., & Raza, S. (2025). Bridging stock price data and technical features for stock forecasting: A comparative study of BiLSTM, CNN & GRU models. In 2025 9th International Symposium on Innovative Approaches in Smart Technologies (ISAS). IEEE. https://doi.org/10.1109/ISAS66241.2025.11101961
12. Hassanpour, H. (2020). Evaluation of deep neural network in directional prediction of forex market. TechRxiv. https://doi.org/10.36227/techrxiv.12946994.v1
13. Hewamalage, H., Bergmeir, C., & Bandara, K. (2021). Recurrent neural networks for time series forecasting: Current status and future directions. International Journal of Forecasting, 37(1), 388–427. https://doi.org/10.1016/j.ijforecast.2020.06.008
14. Hsieh, T.-J., Hsiao, H.-F., & Yeh, W. (2011). Forecasting stock markets using wavelet transforms and recurrent neural networks: An integrated system based on artificial bee colony algorithm. Applied Soft Computing, 11(2), 2510–2525. https://doi.org/10.1016/j.asoc.2010.09.007
15. Hochreiter, S., & Schmidhuber, J. (1997). Long short-term memory. Neural Computation, 9(8), 1735–1780. https://doi.org/10.1162/neco.1997.9.8.1735
16. Hyndman, R. J., & Athanasopoulos, G. (2018). Forecasting: Principles and practice (2nd ed.). OTexts. Katterbauer, K., Syed, H., Cleenewerck, L., & Genç, S. Y. (2022). Robo-sukuk pricing for Chinese equities. Borsa İstanbul Review, 22(5), 854–860. https://doi.org/10.1016/j.bir.2022.06.002
17. Kehinde, T. O., Khan, W. A., & Chung, S.-H. (2023). Financial market forecasting using RNN, LSTM, BiLSTM, GRU and transformer-based deep learning algorithms. In Proceedings of the IEOM International Conference on Smart Mobility and Vehicle Electrification (pp. 393–395).
18. Liu, Y., Cao, C., Huang, W., & Hao, S. (2021). A deep neural network based model for stock market prediction. In 2021 IEEE 2nd International Conference on Big Data, Artificial Intelligence and Internet of Things Engineering (ICBAIE) (pp. 320–323). IEEE. https://doi.org/10.1109/ICBAIE52039.2021.9390010
19. Metlek, S. (2022). Forecasting of Dow Jones sukuk index prices using artificial intelligence systems. Economic Computation and Economic Cybernetics Studies and Research, 56(1), 107–124. https://doi.org/10.24818/18423264/56.1.22.07
20. Mursyada, A., & Swandari, F. (2021). Sukuk performance using comparison of return and risk through calculation of market price and fair price. International Journal of Emerging Issues in Islamic Studies, 1(2), 47–59. https://doi.org/10.31098/ijeiis.v1i2.725
21. Paltrinieri, A., Hassan, M. K., Bahoo, S., & Khan, A. (2023). A bibliometric review of sukuk literature. International Review of Economics & Finance, 86, 897–918. https://doi.org/10.1016/j.iref.2019.04.004
22. Sezgin, F. H., Algorabi, Ö., Sart, G., & Güler, M. (2025). Hyperparameter-optimized RNN, LSTM, and GRU models for airline stock price prediction: A comparative study on THYAO and PGSUS. Symmetry, 17(11), 1905. https://doi.org/10.3390/sym17111905
23. Schuster, M., & Paliwal, K. K. (1997). Bidirectional recurrent neural networks. IEEE Transactions on Signal Processing, 45(11), 2673–2681. https://doi.org/10.1109/78.650093
24. Shaikh, Z. M., & Ramadass, S. (2024). Unveiling deep learning powers: LSTM, BiLSTM, GRU, BiGRU, RNN comparison. Indonesian Journal of Electrical Engineering and Computer Science, 35(1), 263–273. https://doi.org/10.11591/ijeecs.v35.i1.pp263-273
25. Sherstinsky, A. (2020). Fundamentals of recurrent neural network (RNN) and long short-term memory (LSTM) network. Physica D: Nonlinear Phenomena, 404, 132306. https://doi.org/10.1016/j.physd.2019.132306
26. Siami-Namini, S., Tavakoli, N., & Namin, A. S. (2019). A comparison of ARIMA and LSTM in forecasting time series. In 2018 17th IEEE International Conference on Machine Learning and Applications (ICMLA) (pp. 1394–1401). IEEE. https://doi.org/10.1109/ICMLA.2018.00227
27. Solikhun, Pujiastuti, L., & Wahyudi, M. (2021). ANN: Predicting of state retail sukuk based on region in Indonesia. Journal of Physics: Conference Series, 1830(1), 012021. IOP Publishing.
28. Usmani, M. T. (2007). Sukuk and their contemporary applications [Konferans sunumu]. AAOIFI Shari’a Council Meeting, Saudi Arabia.
29. Velarde, O., Yao, L., & Zhang, C. (2025). Short-term forecasting with gated recurrent units: An empirical evaluation. arXiv. https://arxiv.org/abs/2504.18185
30. Vishwanath, S. R., & Azmi, S. (2009). An overview of Islamic sukuk bonds. The Journal of Structured Finance, 14(4), 58–67. https://doi.org/10.3905/JSF.2009.14.4.058
31. Wang, J., Du, Y., & Wang, J. (2020). LSTM-based long-term energy consumption prediction with periodicity. Energy, 197, 117197. https://doi.org/10.1016/j.energy.2020.117197
32. Yang, M., & Wang, J. (2022). Adaptability of financial time series prediction based on BiLSTM. Procedia Computer Science, 199, 18–25. https://doi.org/10.1016/j.procs.2022.01.003
33. Zegarra, F. C., Vargas-Machuca, J., & Coronado, A. M. (2023). A comparative study of CNN, LSTM, BiLSTM, and GRU architectures for tool wear prediction in milling processes. Journal of Machine Engineering, 23(4), 122–136. https://doi.org/10.36897/jme/174019
34. Zhang, D., & Kabuka, M. R. (2018). Combining weather condition data to predict traffic flow: A GRU based deep learning approach. IET Intelligent Transport Systems, 12(7), 565–571. https://doi.org/10.1049/iet-its.2017.0313
35. Zhang, G. P. (2003). Time series forecasting using a hybrid ARIMA and neural network model. Neurocomputing, 50, 159–175. https://doi.org/10.1016/S0925-2312(01)00702-0
36. Zulkhibri, M. (2015). A synthesis of theoretical and empirical research on sukuk. Borsa Istanbul Review, 15(4), 237–248. https://doi.org/10.1016/j.bir.2015.06.003