Does Bigger Beat Simpler? Multilayer Perceptrons versus Recurrent Networks in Trade Export Price Forecasting
DOI:
https://doi.org/10.29207/8szpka26Keywords:
KOTRA, big data, Ministry of Trade, Industry and Energy, deep learning, multilayer perceptronAbstract
South Korea’s trade agency, KOTRA, publishes annual trade data under the Ministry of Trade, Industry and Energy that reports gross domestic product (GDP), customs tariffs, an ease-of-business score, and the previous- and current-year price for each traded item, broken down by item and by country. The sheer volume of items and years this data accumulates over makes it hard to extract a meaningful signal by hand, and manual analysis every year is costly in both time and expertise. This paper implements and evaluates a multilayer-perceptron (MLP) model that learns from a single year of this large-scale data to forecast next-year item-level export prices at low computational and human cost. The deep-learning model is first compared against a conventional multivariate-regression model, where it shows a statistically better prediction error and shorter training time. Because export-price data plausibly carries time-series structure, the same task is also modeled with a fully connected MLP and with a recurrent neural network (RNN). While the recurrent model generalizes somewhat better to unseen data, the MLP achieves the stronger overall prediction accuracy on this single-year dataset. Given a longer time series in future work, a recurrent or Transformer-based deep-learning model is expected to improve export-price prediction further.
References
[1] Korea International Trade Association, “K-stat trade statistics,” [Internet], https://stat.kita.net.
[2] B. M. Lee, H. J. Jeong, and K. S. Park, “An influence of the fourth industrial revolution on international trade and countermeasure strategies to promote export in Korea,” Korea Trade Review, vol. 42, no. 3, pp. 1–24, 2017.
[3] S. H. Nam, “Comparison of long-term forecasting performance of export growth rate using time series analysis models and machine learning analysis,” Korea Trade Review, vol. 46, no. 6, pp. 191–209, 2021.
[4] The 9th Public Data Utilization BI Contest, [Internet], http://www.datacontest.kr.
[5] S. Weisberg, Applied Linear Regression. John Wiley & Sons, p. 47, 2005.
[6] Y. LeCun, Y. Bengio, and G. Hinton, “Deep learning,” Nature, vol. 521, no. 7553, pp. 436–444, 2015.
[7] A. Keck, A. Raubold, and A. Truppia, “Forecasting international trade: A time series approach,” OECD Journal: Journal of Business Cycle Measurement and Analysis, vol. 2009, no. 2, pp. 157–176, 2010.
[8] A. W. Veenstra and H. E. Haralambides, “Multivariate autoregressive models for forecasting seaborne trade flows,” Transportation Research Part E: Logistics and Transportation Review, vol. 37, no. 4, pp. 311–319, 2001.
[9] E. S. Silva and H. Hassani, “On the use of singular spectrum analysis for forecasting US trade before, during and after the 2008 recession,” International Economics, vol. 141, pp. 34–49, 2015.
[10] M. E. Torres, M. A. Colominas, G. Schlotthauer, and P. Flandrin, “A complete ensemble empirical mode decomposition with adaptive noise,” in 2011 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), IEEE, 2011.
[11] H. Lu, X. Ma, K. Huang, and M. Azimi, “Carbon trading volume and price forecasting in China using multiple machine learning models,” Journal of Cleaner Production, vol. 249, art. 119386, 2020.
[12] S. Ioffe and C. Szegedy, “Batch normalization: Accelerating deep network training by reducing internal covariate shift,” in International Conference on Machine Learning, PMLR, 2015.
[13] S. Hochreiter and J. Schmidhuber, “Long short-term memory,” Neural Computation, vol. 9, no. 8, pp. 1735–1780, 1997.
[14] Analytics Vidhya, “Feature scaling for machine learning: Understanding the difference between normalization vs. standardization,” [Internet], https://www.analyticsvidhya.com/blog/2020/04/feature-scaling-machine-learning-normalization-standardization/.
[15] N. Srivastava, G. Hinton, A. Krizhevsky, I. Sutskever, and R. Salakhutdinov, “Dropout: A simple way to prevent neural networks from overfitting,” Journal of Machine Learning Research, vol. 15, no. 1, pp. 1929–1958, 2014.
[16] R. Genuer, J.-M. Poggi, and C. Tuleau-Malot, “Variable selection using random forests,” Pattern Recognition Letters, vol. 31, no. 14, pp. 2225–2236, 2010.
[17] C. Molnar, “Permutation feature importance,” [Internet], https://christophm.github.io/interpretable-ml-book/feature-importance.html.
[18] K. Kira and L. A. Rendell, “A practical approach to feature selection,” in Machine Learning Proceedings 1992, Morgan Kaufmann, pp. 249–256, 1992.
[19] J. W. Tukey, Exploratory Data Analysis. Addison-Wesley, 1977.
[20] A. Krizhevsky, I. Sutskever, and G. E. Hinton, “Imagenet classification with deep convolutional neural networks,” Communications of the ACM, vol. 60, no. 6, pp. 84–90, 2017.
[21] Y. LeCun, L. Bottou, Y. Bengio, and P. Haffner, “Gradient-based learning applied to document recognition,” Proceedings of the IEEE, vol. 86, no. 11, pp. 2278–2324, 1998.
[22] A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, Ł. Kaiser, and I. Polosukhin, “Attention is all you need,” Advances in Neural Information Processing Systems, 2017.
Downloads
Published
Issue
Section
License
Copyright (c) 2025 Journal of Systems Engineering and Information Technology (JOSEIT)
This work is licensed under a Creative Commons Attribution 4.0 International License.
Authors who publish with this journal agree to the following terms:
- Authors retain copyright and grant the journal right of first publication with the work simultaneously licensed under Creative Commons Attribution 4.0 International License that allows others to share the work with an acknowledgment of the work's authorship and initial publication in this journal.
- Authors are able to enter into separate, additional contractual arrangements for the non-exclusive distribution of the journal's published version of the work (e.g., post it to an institutional repository or publish it in a book), with an acknowledgment of its initial publication in this journal.
- Authors are permitted and encouraged to post their work online (e.g., in institutional repositories or on their website) prior to and during the submission process, as it can lead to productive exchanges, as well as earlier and greater citation of published work (Refer to The Effect of Open Access).
