BOR SEKTÖRÜNDE AUTOML VE AĞIRLIKLANDIRMA YÖNTEMİ İLE TALEP TAHMİNİ UYGULAMASI

Abstract

Bu çalışmada, bir kurumun sevkiyat verileri kullanılarak talebi doğru tahmin etmeyi amaçlayan bir model önerisi sunulmuştur. AutoML (Otomatik Makine Öğrenmesi) yaklaşımı kapsamında AutoTS (Otomatik Zaman Serisi) kütüphanesi ile yedi farklı zaman serisi modeli test edilmiş ve performansları çeşitli hata metrikleriyle değerlendirilmiştir. Modelleme sürecinde ürün adı, torbalama türü ve şekli esas alınarak oluşturulan 5 farklı grup için ayrı tahminler yapılmıştır. Tahmin sonuçları ağırlıklandırma yöntemiyle birleştirilerek nihai tahminlere ulaşılmıştır. Çalışma, AutoML temelli zaman serisi tahminlerinin karar destek süreçlerinde etkin bir biçimde kullanılabileceğini göstermektedir. Ayrıca önerilen ağırlıklı modelleme yaklaşımı, dinamik yapısıyla tahmin doğruluğunu artırmaya katkı sağlamaktadır. Kurumun birden fazla ürünü için yapılan tahminler değerlendirilmiş; ancak bu makalede bir ürün üzerinden dört çeyrek dönemlik tahminler ile gerçekleşen değerler karşılaştırılmıştır. Seçilen ürün için çeyrek sonlarında sırasıyla %33, %-82, %-15 ve %0,63 oranlarında sapmalar gözlemlenmiştir. Sonuçlar, modelin bazı dönemlerde sapmalar gösterdiğini, bazı dönemlerde tahmin doğruluğu yüksek olduğunu göstermektedir. 

Keywords

• Talep Tahmini
• Zaman Serisi Analizi
• AutoML
• AutoTS
• Lojistik
• Makine Öğrenmesi

Demand Forecasting Application Using Automl and Weighting Method

Abstract

In this study, a model is proposed to accurately forecast demand based on the shipment data of a public institution. Within the scope of the AutoML (Automated Machine Learning) approach, seven different time series models were tested using the AutoTS (Automated Time Series) library, and their performances were evaluated using various error metrics. During the modeling process, five different groups were created based on variables such as product name, packaging type, and shape, and separate forecasts were generated for each group. The forecasting results were then combined using a weighting method to obtain final predictions. The study demonstrates that AutoML-based time series forecasting can be effectively used in decision support processes. Furthermore, the proposed weighted modeling approach contributes to improving forecast accuracy with its dynamic structure. Although forecasts were generated for multiple products, this paper compares forecasted and actual values for a single product over four quarters. For the selected product, deviations of 33%, -82%, -15%, and 0.63% were observed at the end of each quarter, respectively. The results show that while the model exhibited significant deviations in some periods, it achieved high forecast accuracy in others.

Keywords

• Demand Forecasting
• Time Series Analysis
• AutoML
• AutoTS
• Logistics
• Machine Learning

References

1. Alpay, H., & Yüzügüllü, B. (2005). Eskişehir ve çevresindeki üretim işletmelerinde üretim yönetimi uygulamaları. Eskişehir Osmangazi Üniversitesi.
2. Cerqueira, V., Torgo, L., & Soares, C. (2022). A case study comparing machine learning with statistical methods for time series forecasting: Size matters. Journal of Intelligent Information Systems, 59(2), 415–433. https://doi.org/10.1007/s10844-022-00713-9
3. Contreras, J., Espinola, R., Nogales, F. J., & Conejo, A. J. (2003). ARIMA models to predict next-day electricity prices. IEEE Transactions on Power Systems, 18(3), 1014–1020. https://doi.org/10.1109/TPWRS.2002.804943
4. Dietterich, T. G. (2000). Ensemble methods in machine learning. In J. Kittler & F. Roli (Eds.), Multiple classifier systems (pp. 1–15). Springer. https://doi.org/10.1007/3-540-45014-9_1
5. Fox, J., & Weisberg, S. (2018). Time-series regression and generalized least squares in R: An appendix to An R companion to applied regression (3rd ed.). SAGE Publications
6. He, X., Zhao, K., & Chu, X. (2019). AutoML: A Survey of the State-of-the-Art. arXiv. https://doi.org/10.48550/arXiv.1908.00709
7. Heckbert, P. S. (1998). Fourier transforms and the fast Fourier transform (FFT) algorithm. (Revize edilmiş basım). [Yayınlanmamış çalışma veya Teknik Rapor]. Carnegie Mellon University.
8. Hyndman, R. J., & Athanasopoulos, G. (2018). Forecasting: Principles and practice (2nd ed.). OTexts. https://otexts.com/fpp2/

9. Hyndman, R. J., & Koehler, A. B. (2006). Another look at measures of forecast accuracy. International Journal of Forecasting, 22(4), 679–688. https://doi.org/10.1016/j.ijforecast.2006.03.001
10. Makridakis, S., Spiliotis, E., & Assimakopoulos, V. (2018). The M4 competition: Results, findings, conclusion and way forward. International Journal of Forecasting, 34(4), 802–808. https://doi.org/10.1016/j.ijforecast.2018.06.001
11. McCullagh, P., & Nelder, J. A. (1989). Generalized linear models (2nd ed.). Chapman and Hall/CRC. https://doi.org/10.1201/9780203753736
12. Meisenbacher, S., Turowski, M., Phipps, K., Rätz, M., Müller, D., Hagenmeyer, V., & Mikut, R. (2022). Review of automated time series forecasting pipelines. Sensors, 22(19), 7247. https://doi.org/10.3390/s22197247
13. Montero-Manso, P., Athanasopoulos, G., Hyndman, R. J., & Talagala, T. S. (2020). FFORMA: Feature-based forecast model averaging. International Journal of Forecasting, 36(1), 86–92. https://doi.org/10.1016/j.ijforecast.2019.02.011
14. Opitz, D., & Maclin, R. (1999). Popular ensemble methods: An empirical study. Journal of Artificial Intelligence Research, 11, 169–191. https://doi.org/10.1613/jair.614
15. Park, Y., Lee, J., & Park, W. (2021). Self-weighted ensemble method to adjust the influence of individual models based on reliability. In 2021 International Conference on Information and Communication Technology Convergence (ICTC) (pp. 1195–1197). IEEE. https://doi.org/10.48550/arXiv.2104.04120
16. Pedregosa, F., Varoquaux, G., Gramfort, A., Michel, V., Thirion, B., Grisel, O., Blondel, M., Prettenhofer, P., Weiss, R., Dubourg, V., Vanderplas, J., Passos, A., Cournapeau, D., Brucher, M., Perrot, M., & Duchesnay, É. (2011). Scikit-learn: Machine Learning in Python. Journal of Machine Learning Research, 12, 2825–2830. https://doi.org/10.5555/1953048.2078195
17. Qiu, X., Hu, J., Zhou, L., & Wu, X. (2022). TFB: Towards comprehensive and fair benchmarking of time series forecasting methods. IEEE Transactions on Knowledge and Data Engineering, 34(8), 3918–3932. https://doi.org/10.48550/arXiv.2403.20150
18. Taylor, S. J., & Letham, B. (2018). Forecasting at scale. The American Statistician, 72(1), 37–45. https://doi.org/10.1080/00031305.2017.1380080
19. Top, S., & Yılmaz, C. (2009). Üretim yönetimi. İstanbul Ticaret Üniversitesi.
20. Triebe, O., Hewamalage, H., & Seeger, M. (2021). NeuralProphet: Explainable forecasting at scale. arXiv preprint arXiv:2101.02118. https://doi.org/10.48550/arXiv.2111.1539
21. Wang, C., Chen, X., Wu, C., & Wang, H. (2022). AutoTS: Automatic Time Series Forecasting Model Design Based on Two-Stage Pruning. IEEE Transactions on Knowledge and Data Engineering, 34(8), 3848–3862. https://doi.org/10.48550/arXiv.2203.14169