VUCA KOŞULLARINDA TÜRKİYE’NİN SERAMİK SOFRA EŞYALARI İHRACATINDA CİDDİ DARALMA RİSKİ: AÇIKLANABİLİR MAKİNE ÖĞRENMESİ YAKLAŞIMI

Abstract

Bu çalışma, Türkiye’nin seramik sofra eşyaları ihracatında gelecek yıl ortaya çıkabilecek ciddi daralma riskini incelemektedir. Analiz, 2003–2024 döneminde HS 691110 ve HS 691200 ürün grupları için 168 hedef ülkeyi kapsayan ülke–ürün–yıl verisine dayanmaktadır. Ciddi daralma, bir pazara yapılan ihracatın izleyen yılda en az %30 azalması olarak tanımlanmıştır. Lojistik regresyon, Random Forest, HistGradientBoosting ve XGBoost modelleri zaman sırasını koruyan eğitim, doğrulama ve test dönemlerinde karşılaştırılmıştır. Ortak 975 gözlemli test örnekleminde geliştirilmiş XGBoost modeli 0,6655 ROC-AUC, 0,4247 PR-AUC ve 0,6260 dengeli doğruluk değerine ulaşmıştır. Yeni model, ilk modele göre daha az yanlış alarm üretmiş; buna karşılık bazı gerçek daralma vakalarını kaçırmıştır. Açıklanabilirlik analizi, geçmiş ihracat eğilimi, oynaklık, birim değer hareketleri, hedef ülkedeki enflasyon ve büyüme ile pazar yoğunlaşmasının tahminlerde öne çıktığını göstermiştir. Bulgular, modelin kesin bir tahmin sistemi olmaktan çok, yakından izlenmesi gereken hedef pazarları belirlemeye yardımcı olan bir erken uyarı aracı olarak kullanılabileceğine işaret etmektedir.

Keywords

• VUCA
• seramik sofra eşyaları
• ihracat daralması
• pazar riski
• XGBoost
• erken uyarı

PREDICTING SEVERE EXPORT CONTRACTION RISK IN TÜRKİYE’S CERAMIC TABLEWARE TRADE UNDER VUCA CONDITIONS: AN EXPLAINABLE MACHINE LEARNING APPROACH

Abstract

This study examines the risk of severe export contraction in Türkiye’s ceramic tableware trade. The analysis uses country–product–year data covering 168 destination markets for HS 691110 and HS 691200 during 2003–2024. Severe contraction is defined as a decline of at least 30% in exports to a destination market in the following year. Logistic regression, Random Forest, HistGradientBoosting, and XGBoost models were compared through chronologically separated training, validation, and test periods. On the common test sample of 975 observations, the improved XGBoost model achieved a ROC-AUC of 0.6655, a PR-AUC of 0.4247, and a balanced accuracy of 0.6260. The improved model produced fewer false alarms than the initial model, although it missed some actual contraction cases. The explanation analysis showed that previous export patterns, volatility, unit-value movements, inflation, economic growth, and market concentration were influential in the predictions. The findings suggest that the model is more suitable as an early-warning and market-prioritization tool than as a precise forecasting system.

Keywords

• VUCA
• ceramic tableware
• export contraction
• market risk
• XGBoost
• early warning

References

1. [1] Bennett, N., & Lemoine, G. J. (2014). What a difference a word makes: Understanding threats to performance in a VUCA world. Business Horizons, 57(3), 311–317. https://doi.org/10.1016/j.bushor.2014.01.001
2. [2] Handley, K., & Limão, N. (2015). Trade and investment under policy uncertainty: Theory and firm evidence. American Economic Journal: Economic Policy, 7(4), 189–222. https://doi.org/10.1257/pol.20140068
3. [3] Handley, K., & Limão, N. (2017). Policy uncertainty, trade, and welfare: Theory and evidence for China and the United States. American Economic Review, 107(9), 2731–2783. https://doi.org/10.1257/aer.20141419
4. [4] Carballo, J., Handley, K., & Limão, N. (2022). Economic and policy uncertainty: Aggregate export dynamics and the value of agreements. Journal of International Economics, 139, 103661. https://doi.org/10.1016/j.jinteco.2022.103661
5. [5] Bonadio, B., Huo, Z., Levchenko, A. A., & Pandalai-Nayar, N. (2021). Global supply chains in the pandemic. Journal of International Economics, 133, 103534. https://doi.org/10.1016/j.jinteco.2021.103534
6. [6] Ahir, H., Bloom, N., & Furceri, D. (2022). The world uncertainty index (NBER Working Paper No. 29763). National Bureau of Economic Research. https://doi.org/10.3386/w29763
7. [7] Chen, T., & Guestrin, C. (2016). XGBoost: A scalable tree boosting system. In Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (pp. 785–794). New York, NY: Association for Computing Machinery. https://doi.org/10.1145/2939672.2939785
8. [8] Lundberg, S. M., & Lee, S.-I. (2017). A unified approach to interpreting model predictions. In I. Guyon, U. von Luxburg, S. Bengio, H. Wallach, R. Fergus, S. Vishwanathan, & R. Garnett (Eds.), Advances in Neural Information Processing Systems 30 (pp. 4765–4774). Curran Associates, Inc. https://proceedings.neurips.cc/paper_files/paper/2017/hash/8a20a8621978632d76c43dfd28b67767-Abstract.html

9. [9] Besedeš, T., & Prusa, T. J. (2006). Product differentiation and duration of US import trade. Journal of International Economics, 70(2), 339–358. https://doi.org/10.1016/j.jinteco.2005.12.005
10. [10] Almunia, M., Antràs, P., Lopez-Rodriguez, D., & Morales, E. (2021). Venting out: Exports during a domestic slump. American Economic Review, 111(11), 3611–3662. https://doi.org/10.1257/aer.20181853
11. [11] Chatterjee, A., Dix-Carneiro, R., & Vichyanond, J. (2013). Multi-product firms and exchange rate fluctuations. American Economic Journal: Economic Policy, 5(2), 77–110. https://doi.org/10.1257/pol.5.2.77
12. [12] Breiman, L. (2001). Random forests. Machine Learning, 45(1), 5–32. https://doi.org/10.1023/A:1010933404324
13. [13] Friedman, J. H. (2001). Greedy function approximation: A gradient boosting machine. The Annals of Statistics, 29(5), 1189–1232. https://doi.org/10.1214/aos/1013203451
14. [14] Saito, T., & Rehmsmeier, M. (2015). The precision-recall plot is more informative than the ROC plot when evaluating binary classifiers on imbalanced datasets. PLOS ONE, 10(3), e0118432. https://doi.org/10.1371/journal.pone.0118432
15. [15] Brier, G. W. (1950). Verification of forecasts expressed in terms of probability. Monthly Weather Review, 78(1), 1–3. https://doi.org/10.1175/1520-0493(1950)078<0001:VOFEIT>2.0.CO;2
16. [16] Silva, T. C., Wilhelm, P. V. B., & Amancio, D. R. (2024). Machine learning and economic forecasting: The role of international trade networks. Physica A: Statistical Mechanics and its Applications, 649, 129977. https://doi.org/10.1016/j.physa.2024.129977
17. [17] Shi, G., Chen, Z., Luo, W., & Wei, Z. (2024). Cross-border spillover of imported sovereign risk to China: Key factors identification based on XGBoost-SHAP explainable machine learning algorithm. Finance Research Letters, 70, 106307. https://doi.org/10.1016/j.frl.2024.106307
18. [18] Purnell, D., Jr., Etemadi, A., & Kamp, J. (2024). Developing an early warning system for financial networks: An explainable machine learning approach. Entropy, 26(9), 796. https://doi.org/10.3390/e26090796
19. [19] United Nations Statistics Division. (n.d.). UN Comtrade Database [Data set]. Retrieved July 30, 2026, from https://comtradeplus.un.org/
20. [20] World Bank. (n.d.). World Development Indicators [Data set]. Retrieved July 30, 2026, from https://databank.worldbank.org/source/world-development-indicators
21. [21] International Monetary Fund. (n.d.). Global price of Natural gas, EU [PNGASEUUSDM] [Data set]. Retrieved July 30, 2026, from FRED, Federal Reserve Bank of St. Louis: https://fred.stlouisfed.org/series/PNGASEUUSDM
22. [22] Bergmeir, C., Hyndman, R. J., & Koo, B. (2018). A note on the validity of cross-validation for evaluating autoregressive time series prediction. Computational Statistics & Data Analysis, 120, 70–83. https://doi.org/10.1016/j.csda.2017.11.003