Forecasting the Future Values of Turkey's Scrap Iron and Steel Imports with Deep Learning, Machine Learning and Ensemble Learning Methods

Year 2024, , 885 - 908, 30.09.2024

Yunus Emre Gür , Kamil Abdullah Eşidir

https://doi.org/10.29023/alanyaakademik.1497646

Abstract

This study comprehensively evaluates LSTM, MLP, Random Forest, SVM, XGBoost, and linear regression models to forecast Turkey's scrap iron and steel imports. The performance of the models is measured using RMSE, MSE, MAE, MAPE and R² metrics. The LSTM model shows the best forecasting performance, achieving RMSE 0.0387, MSE 0.0014, MAE 0.0297, MAPE 0.1261, and R² 0.9631 in the training set. According to import forecasts for the next 12 months, imports are expected to increase from 773,378,496 USD in April 2024 to 1,239,538,176 USD in March 2025. LIME analysis makes the decision-making processes of the model transparent by explaining which independent variables the model is based on. As a result of the analysis, it is determined that the model attaches great importance to variables such as “PPI” and “Monthly Iron and Steel Imports”, and the impact of these variables on the forecast results is more significant than other independent variables. With this analysis, the impact of each independent variable on the results of the model is visualized and the contribution levels of the variables are evaluated to reveal which features are given more weight by the model.

Keywords

Scrap Iron and Steel Imports, LSTM, Machine Learning, Deep Learning, Ensemble Learning, LIME (Local Interpretable Model-agnostic Explanations) Analysis

Türkiye Hurda Demir Çelik İthalatının Gelecek Değerlerinin Derin Öğrenme, Makine Öğrenmesi ve Topluluk Öğrenme Yöntemleri ile Öngörülmesi

Abstract

Bu çalışma, Türkiye’nin hurda demir çelik ithalatını tahmin etmek için LSTM, MLP, Random Forest, SVM, XGBoost ve Doğrusal Regresyon modellerini kapsamlı bir şekilde değerlendirmektedir. Modellerin performansları RMSE, MSE, MAE, MAPE ve R² metrikleri kullanılarak ölçülmüştür. LSTM modeli, en iyi tahmin performansını göstererek eğitim setinde RMSE 0,0387, MSE 0,0014, MAE 0,0297, MAPE 0,1261 ve R² 0.9631 sonuçlarını elde etmiştir. Gelecek 12 aylık ithalat tahminlerine göre, Nisan 2024’te 773.378.496 USD olan ithalat miktarının Mart 2025'te 1.239.538.176 USD’ye ulaşması beklenmektedir. LIME analizi, modelin hangi bağımsız değişkenlere dayandığını açıklayarak modelin karar verme süreçlerini şeffaf hale getirmektedir. Analiz sonucunda, modelin özellikle “YÜFE” ve “Aylık Demir Çelik İthalatı” gibi değişkenlere yüksek önem verdiği, bu değişkenlerin tahmin sonuçları üzerindeki etkisinin diğer bağımsız değişkenlere göre daha belirgin olduğu tespit edilmiştir. Bu analiz ile her bir bağımsız değişkenin modelin sonuçları üzerindeki etkisi görselleştirilmiş ve değişkenlerin katkı düzeyleri değerlendirilerek modelin hangi özelliklere daha fazla ağırlık verdiği ortaya konulmuştur.

Keywords

Hurda Demir Çelik İthalatı, LSTM, Makine Öğrenmesi, Derin Öğrenme, Topluluk Öğrenme, LIME (Local Interpretable Model-agnostic Explanations) Analizi

References

• Adhikari, G. P. (2022). Interpreting the basic results of multiple linear regression. Scholars' Journal, 5(1), 22-37. https://doi.org/10.3126/scholars.v5i1.55775
• Alanezi, S. T., Kraśny, M. J., Kleefeld, C., & Colgan, N. (2023). Differential diagnosis of prostate cancer grade to augment clinical diagnosis based on classifier models with tuned hyperparameters. Cancers, 16, 2163 https://doi.org/10.20944/preprints202311.1822.v1
• Albayrak, B. A. (2011). Dünya Hurda Hareketleri ve 2020 Türkiye Projeksiyonu. Yüksek Lisans Tezi, İstanbul Teknik Üniversitesi, Fen Bilimleri Enstitüsü, s. 5-11.
• Ampomah, E. K., Nyame, G., Qin, Z., Addo, P. C., Gyamfi, E. O., & Gyan, M. (2021). Stock market prediction with gaussian naïve bayes machine learning algorithm. Informatica, 45(2). https://doi.org/10.31449/inf.v45i2.3407
• Bilgili, M., Keiyinci, S., & Ekinci, F. (2022). One-day ahead forecasting of energy production from run-of-river hydroelectric power plants with a deep learning approach. Scientia Iranica, 29(4), 1838-1852. https://doi.org/10.24200/sci.2022.58636.5825
• Bulum A. Z. (2015). Türkiye demir-çelik sektörü için tahmin modelleri önerisi. Yüksek Lisans Tezi, Karabük Üniversitesi, Fen Bilimleri Enstitüsü, Endüstri Mühendisliği Anabilim Dalı, Karabük.
• Cheng, C., & Tsai, M. (2022). An intelligent time series model based on hybrid methodology for forecasting concentrations of significant air pollutants. Atmosphere, 13(7), 1055. https://doi.org/10.3390/atmos13071055
• Çetin, B., & Filiz, T. (2023). Küresel hurda demir ticareti ilişkilerinin sosyal ağ analizi yöntemiyle değerlendirilmesi. Mehmet Akif Ersoy Üniversitesi İktisadi ve İdari Bilimler Fakültesi Dergisi, 10 (1) , 158-182. DOI: 10.30798/makuiibf.1097376
• Çubuk, M. (2021). Çok kriterli karar verme yöntemleri ile illerin yatırım ortamlarının karşılaştırılması. Ankara: Gazi Kitabevi.
• Dhakal, S., Gautam, Y., & Bhattarai, A. (2020). Exploring a deep lstm neural network to forecast daily pm2.5 concentration using meteorological parameters in kathmandu valley, nepal. Air Quality, Atmosphere & Health, 14(1), 83-96. https://doi.org/10.1007/s11869-020-00915-6
• Ding, J., & Feng, D. (2023). Feature selection of ground motion intensity measures for data‐driven surrogate modeling of structures. Earthquake Engineering & Structural Dynamics, 53(3), 1216-1237. https://doi.org/10.1002/eqe.4068
• Elias, R., Fang, L., & Wahab, M. (2011). Electricity load forecasting based on weather variables and seasonalities: a neural network approach. Icsssm11. https://doi.org/10.1109/icsssm.2011.5959472
• Ferrat, L. A., Goodfellow, M., & Terry, J. R. (2018). Classifying dynamic transitions in high dimensional neural mass models: a random forest approach. PLOS Computational Biology, 14(3), e1006009. https://doi.org/10.1371/journal.pcbi.1006009
• Gujarati, D. N. (2003). Basic econometrics, McGraw Hill, Newyork.
• Güner, Ş. N., & Demir, H. U. (2022). Yapay sinir ağları ve zaman serileri yöntemi ile demir çelik ithalatı tahmini. Sakarya İktisat Dergisi, 11 (3), 389-397.
• Haq, M., Ahmed, A., Khan, I., Gyani, J., Mohamed, A., Attia, E., … & Mangan, P. (2022). Analysis of environmental factors using ai and ml methods. Scientific Reports, 12(1). https://doi.org/10.1038/s41598-022-16665-7
• Jiang, S., Xinyue, S., & Zheng, Z. (2019). Gaussian process-based hybrid model for predicting oxygen consumption in the converter steelmaking process. Processes, 7(6), 352. https://doi.org/10.3390/pr7060352
• Kakade, K., Mishra, A. K., Ghate, K., & Gupta, S. (2022). Forecasting commodity market returns volatility: a hybrid ensemble learning garch‐lstm based approach. Intelligent Systems in Accounting, Finance and Management, 29(2), 103-117. https://doi.org/10.1002/isaf.1515
• Karaev, A. K., Gorlova, O. S., Ponkratov, V. V., Vasyunina, M. L., Masterov, A. I., & Sedova, M. L. (2023). Program-target mechanisms to ensure the fiscal balance of the federal constituent. Emerging Science Journal, 7(5), 1517-1533. https://doi.org/10.28991/esj-2023-07-05-05
• Kulkarni, V., & Sinha, P. K. (2012). Pruning of random forest classifiers: a survey and future directions. 2012 International Conference on Data Science & Engineering (ICDSE). https://doi.org/10.1109/icdse.2012.6282329
• Lee, H., & Sohn, I. S. (2015). Global scrap trading outlook analysis for steel sustainability. Journal of Sustainable Metallurgy, 1(1), 39-52. https://doi.org/10.1007/s40831-015-0007-7
• Lei, B., Liu, Z., & Song, Y. (2021). On stock volatility forecasting based on text mining and deep learning under high‐frequency data. Journal of Forecasting, 40(8), 1596-1610. https://doi.org/10.1002/for.2794
• Li, Y., Jia, Z., Liu, Z., Shao, H., Zhao, W., Liu, Z., … & Wang, B. (2024). Interpretable intelligent fault diagnosis strategy for fixed-wing uav elevator fault diagnosis based on improved cross entropy loss. Measurement Science and Technology, 35(7), 076110. https://doi.org/10.1088/1361-6501/ad3666
• Liu, H., Li, Q., Li, G., & Ding, R. (2020). Life cycle assessment of environmental impact of steelmaking process. Complexity, 2020, 1-9. https://doi.org/10.1155/2020/8863941
• Man, H., Huang, H., Qin, Z., & Li, Z. (2023). Analysis of a sarima-xgboost model for hand, foot, and mouth disease in xinjiang, china. Epidemiology and Infection, 151. https://doi.org/10.1017/s0950268823001905
• Mao, Y., Pranolo, A., Wibawa, A. P., Utama, A. B. P., & Dwiyanto, F. A. (2022). Robust lstm with tuned-pso and bifold-attention mechanism for analyzing multivariate time-series. IEEE Access, 10, 78423-78434. https://doi.org/10.1109/access.2022.3193643
• Meng, F., Weng, K., Shallal, B., Chen, X., & Mourshed, M. (2018). Forecasting algorithms and optimization strategies for building energy management & demand response. Sp 2018. https://doi.org/10.3390/proceedings2151133
• Metlek, S., Kandilli, C., & Kayaalp, K. (2021). Prediction of the effect of temperature on electric power in photovoltaic thermal systems based on natural zeolite plates. International Journal of Energy Research, 46(5), 6370-6382. https://doi.org/10.1002/er.7575
• Monjon, S., & Quirion, P. (2010). How to design a border adjustment for the european union emissions trading system? Energy Policy, 38(9), 5199-5207. https://doi.org/10.1016/j.enpol.2010.05.005
• Moon, J., Ke, F., & Sokolikj, Z. (2020). Automatic assessment of cognitive and emotional states in virtual reality‐based flexibility training for four adolescents with autism. British Journal of Educational Technology, 51(5), 1766-1784. https://doi.org/10.1111/bjet.13005
• Mumcu, Z. (2003). Demir çelik hurda raporu. İstanbul Ticaret Odası Dış Ticaret Şubesi, s. 1-5.
• Mussumeci, E., & Coelho, F. C. (2020). Large-scale multivariate forecasting models for dengue - lstm versus random forest regression. Spatial and Spatio-Temporal Epidemiology, 35, 100372. https://doi.org/10.1016/j.sste.2020.100372
• Muthukrishnan, R., & Kalaivani, S. (2023). Robust weighted support vector regression approach for predictive modeling. Indian Journal of Science and Technology, 16(30), 2287-2296. https://doi.org/10.17485/ijst/v16i30.1180
• Naimi, B., Hamm, N., Groen, T., Skidmore, A. K., & Toxopeus, A. (2013). Where is positional uncertainty a problem for species distribution modelling? Ecography, 37(2), 191-203. https://doi.org/10.1111/j.1600-0587.2013.00205.x
• Naito, M., Takeda, K., & Matsui, Y. (2015). Ironmaking technology for the last 100 years: deployment to advanced technologies from introduction of technological know-how, and evolution to next-generation process. ISIJ International, 55(1), 7-35. https://doi.org/10.2355/isijinternational.55.7
• Naseem, M., Chaudhary, K., Sharma, B. N., & Lal, A. G. (2019). Using ensemble decision tree model to predict student dropout in computing science. 2019 IEEE Asia-Pacific Conference on Computer Science and Data Engineering (CSDE). https://doi.org/10.1109/csde48274.2019.9162389
• Norouzi, M., Collins, M. D., Fleet, D. J., & Kohli, P. (2015). CO₂ forest: improved random forest by continuous optimization of oblique splits. https://doi.org/10.48550/arxiv.1506.06155
• Ozemre, M., & Kabadurmuş, Ö. (2020). A big data analytics based methodology for strategic decision making. Journal of Enterprise Information Management, 33(6), 1467-1490. https://doi.org/10.1108/jeim-08-2019-0222
• Pauna, H., Aula, M., Seehausen, J., Klung, J., Huttula, M., & Fabritius, T. (2020). Optical emission spectroscopy as an online analysis method in industrial electric arc furnaces. Steel Research International, 91(11). https://doi.org/10.1002/srin.202000051
• Polat, M. (2019). Petrol fiyatlarının ve reel efektif döviz kurunun türkiye’nin dış ticaret dengesine etkileri: sınır testi yaklaşımı. Maliye Finans Yazıları, (112), 149-174. https://doi.org/10.33203/mfy.602961
• Roy, D. K., Sarkar, T. K., Kamar, S. S. A., Goswami, T., Muktadir, A., Al-Ghobari, H. M., … & Mattar, M. A. (2022). Daily prediction and multi-step forward forecasting of reference evapotranspiration using lstm and bi-lstm models. Agronomy, 12(3), 594. https://doi.org/10.3390/agronomy12030594
• Ryll, L., & Seidens, S. (2019). Evaluating the performance of machine learning algorithms in financial market forecasting: a comprehensive survey.. https://doi.org/10.48550/arxiv.1906.07786
• Setiawati, F., Ahmad, M. S., & Adiatman, M. (2023). Correlates of dental visits in children with hearing loss: an application of the theory of planned behaviour. International Journal of Paediatric Dentistry, 33(3), 259-268. https://doi.org/10.1111/ipd.13036
• Shin, S., Kwon, M., Kim, S., & So, H. (2023). Prediction of equivalence ratio in combustion flame using chemiluminescence emission and deep neural network. International Journal of Energy Research, 2023, 1-10. https://doi.org/10.1155/2023/3889951
• Si, Y., Nadarajah, S., Zhang, Z., & Xu, C. (2024). Modeling opening price spread of shanghai composite index based on arima-gru/lstm hybrid model. Plos One, 19(3), e0299164. https://doi.org/10.1371/journal.pone.0299164
• Silva, F., Sanz, M., Seixas, J., Solano, E., & Omar, Y. (2020). Perceptrons from memristors. Neural Networks, 122, 273-278. https://doi.org/10.1016/j.neunet.2019.10.013
• Şeker, M., Memmedov, A., Huseyinov, R., & Koçkanat, S. (2017). Power quality measurement and analysis in electric arc furnace for turkish electricity transmission system. Elektronika Ir Elektrotechnika, 23(6). https://doi.org/10.5755/j01.eie.23.6.19691
• Taşdelen İ. (2018). Türkiye’de kur değişimlerinin demir-çelik sektörü dış ticareti üzerine etkisi. Yüksek Lisans Tezi, Dokuz Eylül Üniversitesi, Sosyal Bilimler Enstitüsü, İktisat Anabilim Dalı, İzmir.
• Torres, G. d. C., Roig-Maimó, M. F., Mascaró-Oliver, M., Alcover, E. A., & Mas-Sansó, R. (2022). Understanding how cnns recognize facial expressions: a case study with lime and cem. Sensors, 23(1), 131. https://doi.org/10.3390/s23010131
• Truong, N., Ngo, N., & Pham, A. (2021). Forecasting time-series energy data in buildings using an additive artificial intelligence model for improving energy efficiency. Computational Intelligence and Neuroscience, 2021, 1-12. https://doi.org/10.1155/2021/6028573
• Türkiye Cumhuriyet Merkez Bankası Elektronik Veri Dağıtım Sistemi, (2024), https://evds2.tcmb.gov.tr/index.php?/evds/serieMarket, , Erişim Tarihi: 07.05.2024.
• Türkiye İstatistik Kurumu, (2024), www.tuik.gov.tr, Erişim Tarihi: 09.05.2024.
• Tyass, I., Khalili, T., Rafik, M., Bellat, A., Raihani, A., & Mansouri, K. (2023). Wind speed prediction based on statistical and deep learning models. International Journal of Renewable Energy Development, 12(2), 288-299. https://doi.org/10.14710/ijred.2023.48672
• Wang, X., Yu, Y., Zhao, X., Huang, M., & Zhu, Q. (2023). Integrating field images and microclimate data to realize multi-day ahead forecasting of maize crop coverage using cnn-lstm. International Journal of Agricultural and Biological Engineering, 16(2), 199-206. https://doi.org/10.25165/j.ijabe.20231602.7020
• Wang, Y. (2022). Import and export trade forecasting algorithm based on blockchain security and pso optimized hybrid rvm model. Security and Privacy, 6(2). https://doi.org/10.1002/spy2.218
• Yan, L., Jia, L., Lu, S., Peng, L., & He, Y. (2023). Lstm‐based deep learning framework for adaptive identifying eco‐driving on intelligent vehicle multivariate time‐series data. IET Intelligent Transport Systems, 18(1), 186-202. https://doi.org/10.1049/itr2.12443
• Yılmaz, S., Yücer, H. M., & Fidan, Y. (2013). Socio-economic status, job satisfaction and sector visions of the steelworkers: karabuk case study. Journal of History Culture and Art Research, 2(3), 259. https://doi.org/10.7596/taksad.v2i3.271
• Yu, M., Liu, T., Guan, Z., Sun, Y., Jie, G., Chen, L., … & He, Y. (2022). Salstm: an improved lstm algorithm for predicting the competitiveness of export products. International Journal of Intelligent Systems, 37(9), 6185-6200. https://doi.org/10.1002/int.22839
• Yücekutlu, A. Y., & Sanalan A. T. (2015). Elektrik ark ocaklı çelikhane tesislerinin; hava kirleticileri, emisyon kontrol ve azaltım teknikleri, 6. Ulusal Hava Kirliliği ve Kontrolü Sempozyumu, 7-9 Ekim 2015, İzmir.
• Zafar, M. R., & Khan, N. (2021). Deterministic local interpretable model-agnostic explanations for stable explainability. Machine Learning and Knowledge Extraction, 3(3), 525-541. https://doi.org/10.3390/make3030027
• Zhou, Y., Qi, L., & Xiao, D. (2022). Application of lstm-lightgbm nonlinear combined model to power load forecasting. Journal of Physics: Conference Series, 2294(1), 012035. https://doi.org/10.1088/1742-6596/2294/1/012035