本研究將以優化個案公司的預測庫存量作為主要目的，分別應用樣本平均近似法(Sample average approximation, SAA)、深度學習模型(Deep learning, DL)、整合模型進行預測庫存量研究，將最終分析結果作為個案公司的庫存管理作業模式轉型之參考建議。本研究將以庫存管理相關內容進行文獻蒐集與彙整，如預測庫存量的常見過往處理方法、樣本平均近似法、深度學習與整合模型相關內容進行文獻探討。最後，在應用模型與數據研究方面使用實際個案資料進行模型求解，最終結果將以多個情境狀況進行分析研究。
在查閱過往相關文獻可知，由於在許多研究結果表示，大多數據可能同時包含線性特性和非線性特性，因此在選擇模型時，則可考慮這些特性因素，選擇有利特性之模型可以提升預測準確性。因此，本研究中選用了三種不同的模型進行庫存數據預測，在過往文獻中，樣本平均近似法適合應用在擁有線性特性的情境，深度學習模型適合應用在非線性特性的情境，最後是整合模型可以包含線性特性與非線性特性的應用。在最終分析結果可以發現整合模型在處理複雜數據建模中的重要性和優越性，且整合模型在三種不同的情境當中都有良好的表現。未來如有相關研究亦可以進一步探索和拓展整合模型的應用，針對不同類型的數據集和行業需求，並開發更為靈活和強大的整合模型，將有助於在更廣泛的背景下提升庫存管理和其他領域的數據預測準確性。

This study has aimed to optimize the forecasted inventory levels for the case company by applying Sample Average Approximation (SAA), Deep Learning models, and Hybrid models for inventory prediction analysis. The final analysis results have been intended as a reference for the case company’s transformation of its inventory management operations. The study has gathered and consolidated literature related to inventory management, including common methods for inventory forecasting, SAA, deep learning, and hybrid models. Finally, actual case data has been used for model-solving in the application of models and data research, and the final results have been analyzed under multiple scenarios.
This study has found that the research data may exhibit both linear and nonlinear characteristics, which should be considered when selecting models to improve forecasting accuracy. As a result, three different models have been used in this study for inventory data prediction: the SAA is suitable for linear characteristics, deep learning models are appropriate for nonlinear characteristics, and hybrid models can incorporate both linear and nonlinear characteristics. The analysis results have highlighted the importance and superiority of hybrid models in handling complex data modeling, demonstrating strong performance across all three different scenarios. Future research can further explore and expand the application of hybrid models, tailoring them to different types of datasets and industry needs. By developing more flexible and powerful hybrid models, it will be possible to enhance the accuracy of inventory management and data forecasting in other fields across a broader context.

Alrweili, H., & Fawzy, H. Forecasting crude oil prices using an ARIMA-ANN hybrid model. J Stat Appl Probab, 11(3), 845-855, 2022
Apicella, A., Donnarumma, F., Isgrò, F., & Prevete, R. (2021). A survey on modern trainable activation functions. Neural Networks, 138, 14-32, 2021
Bertsimas, D., Gupta, V., & Kallus, N. (2018). Robust sample average approximation. Mathematical Programming, 171, 217-282, 2018
Ban, G.-Y. and Rudin, C.The big data newsvendor: Practical insights from machine learning. Operations Research, 67(1), 90–108, 2018
Bertsimas, D. and Kallus, N. From predictive to prescriptive analytics. arXiv preprint arXiv:1402.5481, 2014
Bertsimas, D., Gupta, V., & Kallus, N. Robust sample average approximation. Mathematical Programming, 171, 217-282, 2018
Box, G.E.P., Jenkins, G.M., Reinsel, G.C. and Ljung, G.M. Time Series Analysis: Forecasting and Control, 5th edition, John Wiley & Sons, 2015
Deng, C., & Liu, Y. A Deep Learning‐Based Inventory Management and Demand Prediction Optimization Method for Anomaly Detection. Wireless Communications and Mobile Computing, 2021(1), 2021
Hamedianfar, A., Mohamedou, C., Kangas, A., & Vauhkonen, J. Deep learning for forest inventory and planning: a critical review on the remote sensing approaches so far and prospects for further applications. Forestry, 95(4), 451-465, 2022
Ketkar, N., & Ketkar, N. Stochastic gradient descent. Deep learning with Python: A hands-on introduction, 113-132, 2017
Kleywegt, A. J., Shapiro, A., & Homem-de-Mello, T. The sample average approximation method for stochastic discrete optimization. SIAM Journal on optimization, 12(2), 479-502, 2002
Lauren, S., & Harlili, S. D. Stock trend prediction using simple moving average supported by news classification. In 2014 International Conference of Advanced Informatics: Concept, Theory and Application (ICAICTA), 135-139, 2014
Oroojlooyjadid, A., L. V. Snyder, and M. Taka ́c ̌ Applying deep learning to the newsvendor problem. IISE Transactions 52(4), 444–463, 2020
Qi, M., Shi, Y., Qi, Y., Ma, C., Yuan, R., Wu, D., & Shen, Z. J. A practical end-to-end inventory management model with deep learning. Management Science, 69(2), 759-773, 2023
Shepp, L. A., Slepian, D., & Wyner, A. D. On prediction of moving‐average processes. Bell System Technical Journal, 59(3), 367-415, 1980
Shumway, R.H. and Stoffer, D.S. Time Series Analysis and Its Applications: With R Examples. Springer Science & Business Media, 2010
Soshko, O., & Pluksne, N. Inventory Management in Multi Echelon Supply Chain using Sample Average Approximation. Computer Science, 40, 1407-7493, 2009
Suhermi, N., Prastyo, D. D., & Ali, B. Roll motion prediction using a hybrid deep learning and ARIMA model. Procedia computer science, 144, 251-258, 2018
Van der Laan, N., Teunter, R. H., Romeijnders, W., & Kilic, O. A. The data-driven newsvendor problem: Achieving on-target service-levels using distributionally robust chance-constrained optimization. International Journal of Production Economics, 249, 2022
Vo, T. T. B. C., Le, P. H., Nguyen, N. T., Nguyen, T. L. T., & Do, N. H. Demand Forecasting and Inventory Prediction for Apparel Product using the ARIMA and Fuzzy EPQ Model. Journal of Engineering Science & Technology Review, 14(2), 2021