Forecasting Soybean Production in Turkey: A Comparative Analysis of Automated and Traditional Methods

Abstract

Türkiye’s climate and soil are well-suited for the cultivation of oilseed crops, which are of vital importance to various industries and human and animal diets. Among oilseeds, soybeans, a legume, possess a distinctive nutritional profile. While existing research covers soybean production in Türkiye, this study aims to: a) evaluate production levels using different forecasting algorithms to identify the most accurate model, and b) based on the chosen model, forecast future production and assess the current and future entrepreneurial potential of the soybean industry in Türkiye. Soybean production data (1990-2022) from TURKSTAT was divided into training (n=25) and test (n=8) sets for cross-validation. By applying univariate time series methods, including ARIMA, SES, NNAR, MN, and Naive to the training dataset, it was found that ARIMA (1,1,1) performed best according to test set RMSE values. The performance ranking (in terms of RMSE) was as follows: ARIMA (13019) < SES (13888) < Naive (14240) < NNAR (58393) < MN (80418). Notably, for this dataset, the performance of automated processes was relatively worse than that of manual methods, suggesting that relying solely on automated methods may lead to suboptimal forecasting results. These findings underscore the importance of human oversight in the use of automated algorithms for time series forecasting and highlight the need for caution when employing automated methods. The ARIMA (1,1,1) model predicts a flat trend in production from 2023 to 2032, with an initial production volume of 154 516 tonnes and a slight decline to 153 607 tonnes. This predicted stagnation implies that, in the context of economic and population growth, soybean production will fall further behind domestic demand, leading to increased import reliance. These findings are of serious importance to farmers and policymakers alike, as they can assist in the formulation of informed decisions pertaining to resource allocation, crop planning, and market strategies. Local producers may potentially benefit from increased production efficiency, improved competitiveness, and potential revenue growth by catering to both domestic and export markets. Furthermore, an understanding of these trade dynamics can assist stakeholders in identifying potential avenues for collaboration or investment within the Turkish soybean industry. Further analysis of these results is ongoing in order to gain deeper insights into the factors influencing soybean production trends in Türkiye.

Keywords

• Soybean production
• Türkiye
• Time series forecasting
• ARIMA algorithm
• NNAR
• Auto-ARIMA

References

1. Akın, M., Eyduran, S.P., Çelik, Ş., Aliyev, P., Aykol, S., Eyduran, E., (2021). Modeling and forecasting cherry production in Turkey. Journal of Animal & Plant Sciences 31(3): 773–781. https://doi.org/10.36899/JAPS.2021.3.0267
2. Anonymous., (2022d). Yağlı Tohumlar. https://data.tuik.gov.tr/Bulten/DownloadIstatistikselTablo?p=4VOhKR8BrNY4kV5kQy40A4Ds6MrlLde0frCYb2fEpnEWC9xHHViRbOji7MtzOuUp, (Date of access: September 2, 2023).
3. Anonymous., (2022a). FAOSTAT database: Production - crops and livestock products. https://www.fao.org/faostat/en/#data/QCL, (Date of access: May 20, 2024).
4. Anonymous., (2022b). FAOSTAT database: Trade - crops and livestock products. https://www.fao.org/faostat/en/#data/TCL, Date of access: May 20, 2024).
5. Anonymous., (2022c). FAOSTAT database: Detailed trade matrix. https://www.fao.org/faostat/en/#data/TM, (Date of access: May 20, 2024).
6. Arıoğlu, H., (2016). Türkiye’de yağlı tohum ve ham yağ üretimi, sorunlar ve çözüm önerileri. Tarla Bitkileri Merkez Araştırma Enstitüsü Dergisi 25(ÖZEL SAYI-2): 357–368.
7. Bivand, R.S., Pebesma, E., Gomez-Rubio, V., (2013). Applied spatial data analysis with R, second edition. Springer, NY.
8. Box, G.E.P., Jenkins, G.M., (1970). Time series analysis: Forecasting and control. Holden-Day, San Francisco, USA.

9. Gardner, E.S., Mckenzie, Ed., (1985). Forecasting Trends in Time Series. Management Science 31(10): 1237–1246. https://doi.org/10.1287/mnsc.31.10.1237
10. Gujarati, D.N., Porter, D.C., (2009). Basic Econometrics. McGraw-Hill, Boston, USA.
11. Güler, D., Saner, G., Naseri, Z., (2017). Yağlı tohumlu bitkiler ithalat miktarlarının arıma ve yapay sinir ağları yöntemleriyle tahmini. Balkan ve Yakın Doğu Sosyal Bilimler Dergisi 3(1): 60–70.
12. Holt, C.C., (2004). Forecasting seasonals and trends by exponentially weighted moving averages. International Journal of Forecasting 20(1): 5–10. https://doi.org/https://doi.org/10.1016/j.ijforecast.2003.09.015
13. Hyndman, R.J., (2021). Forecasting: Principles and practice. OTexts, Melbourne, Australia.
14. Hyndman, R.J., Khandakar, Y., (2008). Automatic time series forecasting: The forecast package for r. Journal of Statistical Software 27(3): 1–22. https://doi.org/10.18637/jss.v027.i03
15. Hyndman, R.J., Koehler, A., Ord, K., Snyder, R., (2008). Forecasting with exponential smoothing: The state space approach. Springer, Berlin.
16. Kuhn, M., Johnson, K., (2013). Applied predictive modeling. Springer, New York.
17. Massicotte, P., South, A., (2023). Rnaturalearth: World map data from natural earth.
18. Mélard, G., Pasteels, J-M., (2000). Automatic ARIMA modeling including interventions, using time series expert software. International Journal of Forecasting 16(4): 497–508. https://doi.org/https://doi.org/10.1016/S0169-2070(00)00067-4
19. O’Hara-Wild, M., Hyndman, R., Wang, E., (2023b). Feasts: Feature extraction and statistics for time series.
20. O’Hara-Wild, M., Hyndman, R., Wang, E., (2023a). Fable: Forecasting models for tidy time series.
21. Özcan, M., (2023). Ürün raporu SOYA 2023. Tarımsal Ekonomi ve Politika Geliştirme Enstitüsü, Ankara.
22. Pagano, M.C., Miransari, M., (2016). The importance of soybean production worldwide. In Miransari M (ed.) Abiotic and biotic stresses in soybean production. Academic Press, San Diego, 1–26.
23. Pebesma, E., (2018). Simple Features for R: Standardized Support for Spatial Vector Data. The R Journal 10(1): 439–446. https://doi.org/10.32614/RJ-2018-009
24. Pebesma, E.J., Bivand, R., (2005). Classes and methods for spatial data in R. R News 5(2): 9–13.
25. Pratap, A., Gupta, S.K., Kumar, J., Solanki, R.K., (2012). Soybean. In Gupta SK (ed.) Technological innovations in major world oil crops, volume 1: breeding. Springer New York, New York, NY, 293–321.
26. R Core Team., (2022). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria.
27. South, A., (2017). Rnaturalearthdata: World vector map data from natural earth used in ’rnaturalearth’.
28. Tiwari, S.P., (2017). Emerging trends in soybean industry. Soybean Research 15(1): 1–17.
29. Tüfekçi, Ş., (2019). 2019 ar-ge soya fasülyesi raporu. report. Ereğli Ticaret Borsası, Konya.
30. Uçum, İ., (2016). ARIMA modeli ile Türkiye soya üretim ve ithalat projeksiyonu. Tarım Ekonomisi Araştırmaları Dergisi 2(1): 24–31.
31. Wang, E., Cook, D., Hyndman, R.J., (2020). A new tidy data structure to support exploration and modeling of temporal data. Journal of Computational and Graphical Statistics 29(3): 466–478. https://doi.org/10.1080/10618600.2019.1695624
32. Wickham, H., Averick, M., Bryan, J., Chang, W., McGowan, L.D., François, R., Grolemund, G., Hayes, A., Henry, L., Hester, J., et al., (2019). Welcome to the tidyverse. Journal of Open Source Software 4(43): 1686. https://doi.org/10.21105/joss.01686