Petrol Ürünleri İthalatının Tahmininde Geleneksel, Hibrit ve Meta-Sezgisel Yaklaşımlar: SARIMA, RF-LSTM ve DE-HW Modellerinin Karşılaştırılması

Öz

Bu çalışma, Türkiye’nin 2030 Sürdürülebilir Kalkınma Amaçları doğrultusunda ekonomik ve çevresel sürdürülebilirlik planlamalarına katkı sağlamak amacıyla, 27.10 GTİP kodlu petrol ürünleri ithalat miktarını 2030 yılına kadar tahmin etmeyi ve farklı tahmin modellerinin performanslarını karşılaştırmayı amaçlamaktadır. 2007–2026 dönemine ait aylık ithalat verileri kullanılarak Mevsimsel Otoregresif Entegre Hareketli Ortalama (SARIMA), Diferansiyel Evrim algoritmasıyla optimize edilmiş Holt-Winters (DE-HW) ve derin öğrenme tabanlı Hibrit Rastgele Orman–Kısa Uzun Vadeli Bellek (RF-LSTM) modelleri uygulanmıştır. Model performansları Kök Ortalama Kare Hata (RMSE), Ortalama Kare Hata (MSE), Ortalama Mutlak Hata (MAD) ve Ortalama Mutlak Yüzde Hata (MAPE) ölçütleriyle değerlendirilmiştir. Bulgular, Hibrit RF-LSTM modelinin eğitim ve doğrulama setlerinde sırasıyla %7 ve %5 MAPE değerleriyle en yüksek tahmin performansını sağladığını göstermiştir. Elde edilen 2030 tahminleri, enerji ithalatı bütçe yönetimi, lojistik planlama ve stratejik rezerv kapasitesinin belirlenmesine yönelik politika kararları için veri temelli bir çerçeve sunmaktadır.

Anahtar Kelimeler

• Holt-Winters
• Diferansiyel Evrim Algoritması
• Rastgele Orman
• LSTM
• SARIMA

Advanced Time Series Approaches in Petroleum Product Import Modeling: Comparing Hybrid RF-LSTM, SARIMA, and DE-HW Models

Öz

This study aims to contribute to Türkiye’s economic and environmental sustainability planning in line with the 2030 Sustainable Development Goals by forecasting the import volume of petroleum products classified under HS Code 27.10 through 2030 and comparing the performance of different forecasting models. Using monthly import data for 2007–2026, Seasonal Autoregressive Integrated Moving Average (SARIMA), Holt-Winters optimized by the Differential Evolution algorithm (DE-HW), and a deep learning-based Hybrid Random Forest–Long Short-Term Memory (RF-LSTM) model were applied. Model performances were evaluated using Root Mean Square Error (RMSE), Mean Square Error (MSE), Mean Absolute Deviation (MAD), and Mean Absolute Percentage Error (MAPE). The results showed that the Hybrid RF-LSTM model achieved the highest forecasting performance, with MAPE values of 7% and 5% for the training and validation sets, respectively. The forecasts obtained for 2030 provide a data-driven framework for policy decisions concerning energy import budgeting, logistics planning, and strategic reserve capacity determination.

Anahtar Kelimeler

• Holt-Winters
• Differential Evolution Algorithm
• Random Forest (RF)
• Long Short-Term Memory (LSTM)
• SARIMA

Kaynakça

1. 1. Kökyıldırım, A. & Antmen, Z.F. (2024). Determination of financial performance in electric power enterprises with entropy weighted WASPAS analysis method. Çukurova Üniversitesi Mühendislik Fakültesi Dergisi, 39(3), 849-860.
2. 2. Erkök, B. & Kütük, Y. (2023). Dependency on imported energy in Turkey: Input-output analysis. Marmara Üniversitesi İktisadi ve İdari Bilimler Dergisi, 45(1), 47-70.
3. 3. Ürkmez, İ. & Okyar, M.C. (2022). The effect of renewable energy on energy import dependence: An empirical analysis in Turkey. Siyasal: Journal of Political Sciences, 31(2), 443-462.
4. 4. Özlale, Ü. & Pekkurnaz, D.J.E.P. (2010). Oil prices and current account: A structural analysis for the Turkish Economy. 38(8), 4489-4496.
5. 5. Yalta, A.Y. & Yalta, A.T. (2017). Dependency on imported oil and its effects on current account. Energy Sources, Part B: Economics, Planning, Policy, 12(10), 859-867.
6. 6. Tekin, P. (2022). Çukurova Bölgesi için kısa vadeli yapay zeka tabanlı rüzgar güç tahmini. Çukurova Üniversitesi Mühendislik Fakültesi Dergisi, 37(4), 1143-1154.
7. 7. Ediger, V.Ş. & Berk, I. (2011). Crude oil import policy of Turkey: Historical analysis of determinants and implications since 1968. Energy Policy, 39(4), 2132-2142.
8. 8. Solak, A.O. (2013). Türkiye'nin toplam petrol talebi ve ulaştırma sektörü petrol talebinin Arima Modeli ile tahmin edilmesi. Süleyman Demirel Üniversitesi İktisadi ve İdari Bilimler Fakültesi Dergisi, 18(3), 131-142.

9. 9. Melikoglu, M. (2014). Demand forecast for road transportation fuels including gasoline, diesel, LPG, bioethanol and biodiesel for Turkey between 2013 and 2023. Renewable Energy, 64, 164-171.
10. 10. Sahraei, M.A., Duman, H., Codur, M.Y. & Eyduran, E. (2021). Prediction of transportation energy demand: multivariate adaptive regression splines. Energy, 224, 120090.
11. 11. Ağbulut, Ü. (2022). Forecasting of transportation-related energy demand and CO2 emissions in Turkey with different machine learning algorithms. Sustainable Production Consumption, 29, 141-157.
12. 12. Hoxha, J., Codur, M.Y., Mustafaraj, E., Kanj, H. & El Masri, A. (2023). Prediction of transportation energy demand in Türkiye using stacking ensemble models: Methodology and comparative analysis. Applied Energy, 350, 121765.
13. 13. Kayacı Çodur, M. (2024). Ensemble machine learning approaches for prediction of Türkiye's energy Demand. Energies, 17(1), 74.
14. 14. Bakanlığı, T.C.T. (2026). TAREKS gümrük tarife istatistik pozisyonu (GTİP) sorgulama ekranı. https://uygulama.gtb.gov.tr/Tara/, Erişim tarihi: 15.05.2026.
15. 15. Centre-ITC, U.T.M.-I.T. (2026), Trade Map - International Trade Statistics. https://www.trademap.org/, Erişim tarihi: 05.04.2026.
16. 16. Makoni, T. & Chikobvu, D. (2023). Assessing and forecasting the long-term impact of the global financial crisis on manufacturing sales in South Africa. Economies, 11(6), 158.
17. 17. Nwokike, C.C. & Okereke, E.W. (2021). Comparison of the performance of the SANN, SARIMA and ARIMA models for forecasting quarterly GDP of Nigeria. Asian Research Journal of Mathematics, 17(3), 1-20.
18. 18. Hanke, J.E. & Wichern, D.W. (2009). Business forecasting. Pearson Education, Last Page.
19. 19. Makoni, T. & Chikobvu, D.J.E. (2023). Assessing and forecasting the long-term impact of the global financial crisis on manufacturing sales in South Africa. Economies, 11(6), 158.
20. 20. Dritsakis, N. & Klazoglou, P. (2018). Forecasting unemployment rates in USA using Box-Jenkins methodology. International Journal of Economics Financial Issues 8(1), 9.
21. 21. Chand, S. & Kamal, S. (2014). Mixed portmanteau test for diagnostic checking of time series models. Journal of Applied Mathematics, 2014(1), 545413.
22. 22. Kibona, S.E. & Mbago, M.C. (2018). Forecasting wholesale prices of maize in Tanzania using Arima Model. General Letters in Mathematics, 4(3), 131-141.
23. 23. Ünal, Y.Z., Al-Turki, U.M., Zaim, S., Demirel, Ö.F. & Görener, R. (2015). Developing spreadsheet models of Holt-Winter methods and solving with Microsoft Excel solver and differential evaluation technique: An application to tourism sector. In 2015 International Conference on Industrial Engineering and Operations Management (IEOM), IEEE, 1-8.
24. 24. Eltaeib, T. & Mahmood, A. (2018). Differential evolution: A survey and analysis. Applied Sciences, 8(10), 1945.
25. 25. Zhang, G.P. (2003). Time series forecasting using a hybrid ARIMA and neural network model. Neurocomputing, 50, 159-175.
26. 26. Breiman, L. (2001). Random forests. Machine learning, 45(1), 5-32.
27. 27. Akça, Ş. & Şanlı, F.B. (2022). Kıyı çizgisinin makine öğrenmesi tabanlı mean-shift, rastgele orman ve destek vektör makinaları ile belirlenmesi. Çukurova Üniversitesi Mühendislik Fakültesi Dergisi, 37(3), 835-842.
28. 28. Ran, X., Shan, Z., Fang, Y. & Lin, C. (2019). An LSTM-based method with attention mechanism for travel time prediction. Sensors, 19(4), 861.
29. 29. Oshiro, T.M., Perez, P.S. & Baranauskas, J.A. (2012). How many trees in a random forest?, In International Workshop on Machine Learning and Data Mining in Pattern Recognition, Springer, 154-168.
30. 30. Probst, P., Wright, M.N. & Boulesteix, A.L. (2019). Hyperparameters and tuning strategies for random forest. Wiley Interdisciplinary Reviews: Data Mining Knowledge Discovery, 9(3), e1301.
31. 31. Aree, S. & Tooprakai, S. (2025). Maintenance decision modeling for solar photovoltaic performance prediction using RMSE and MAPE methods. In 2025 International Technical Conference on Circuits/Systems, Computers, and Communications (ITC-CSCC), IEEE, 1-6.