本研究聚焦於營運決策中的兩階段報童問題，提出一套統一式條件生成模型，以建構精確的未來需求分布並支援補貨決策。為因應不同補貨時點所面臨的變長需求序列問題，本文設計結合條件變分自編碼器（Conditional Variational Autoencoder, CVAE）與長短期記憶網路（Long Short-Term Memory, LSTM）之深度生成架構，能根據任意觀測歷史資訊生成滿足條件分布的需求樣本。研究首先於單階段訂價報童問題中驗證 CVAE 於需求重建上的可行性，作為模型基礎能力之驗證，進一步延伸至二階段補貨情境中。實驗採用模擬資料，並與隨機森林模型比較，透過 Wasserstein 距離與 Kolmogorov– Smirnov 統計量評估分布準確性，並分析補貨決策的平均利潤表現。結果顯示所提方法於樣本準確性與決策效益皆具顯著優勢，展現其於需求不確定性建模與營運分析整合應用之潛力。

This study addresses the two-stage newsvendor problem in operations decision-making by proposing a unified conditional generative model to accurately capture the conditional distribution of future demand and support replenishment decisions. To handle the challenge of variable-length demand sequences observed at different replenishment points, we design a deep generative framework that integrates a Conditional Variational Autoencoder (CVAE) with a Long Short-Term Memory (LSTM) network. This model generates conditional demand samples based on any observed historical information. We first validate the feasibility of using CVAE to reconstruct price-driven demand distributions in a single-stage setting, serving as a foundation for model capability, before extending the method to two-stage replenishment scenarios. Simulation experiments compare the proposed model against Random Forest benchmarks. Evaluation metrics include the Wasserstein distance and Kolmogorov– Smirnov (KS) statistic for distribution accuracy, as well as average profit performance for decision quality. Results demonstrate that the proposed method significantly outperforms traditional models in both generative accuracy and downstream decision effectiveness, highlighting its potential for integrating uncertainty modeling with operational analytics.

Aldaihani, M. M. and Darwish, M. A. (2016). Service level enhancement in newsvendor model. Computers & Industrial Engineering, 98:164–170.

Bishop, C. M. and Nasrabadi, N. M. (2006). Pattern recognition and machine learning, volume 4. Springer.

Brégère, M. and Bessa, R. J. (2020). Simulating tariff impact in electrical energy consumption profiles with conditional variational autoencoders. IEEE Access, 8:131949–131966.

Burgess, C. P., Higgins, I., Pal, A., Matthey, L., Watters, N., Desjardins, G., and Lerchner, A. (2018). Understanding disentangling in β-vae. arXiv preprint arXiv:1804.03599.

Contreras, J., Espinola, R., Nogales, F., and Conejo, A. (2003). Arima models to predict next-day electricity prices. IEEE Transactions on Power Systems, 18(3):1014–1020.

DeYong, G. D. (2020). The price-setting newsvendor: review and extensions. International Journal of Production Research, 58(6):1776–1804.

Fang, L., Zeng, T., Liu, C., Bo, L., Dong, W., and Chen, C. (2021). Transformer-based conditional variational autoencoder for controllable story generation. arXiv preprint arXiv:2101.00828.

Gammelli, D., Wang, Y., Prak, D., Rodrigues, F., Minner, S., and Pereira, F. C. (2022). Predictive and prescriptive performance of bike-sharing demand forecasts for inventory management. Transportation Research Part C: Emerging Technologies, 138:103571.

Goodfellow, I., Pouget-Abadie, J., Mirza, M., Xu, B., Warde-Farley, D., Ozair, S., Courville, A., and Bengio, Y. (2020). Generative adversarial networks. Communications of the ACM, 63(11):139–144.

Harsha, P., Natarajan, R., and Subramanian, D. (2021). A prescriptive machine-learning framework to the price-setting newsvendor problem. Informs Journal on Optimization, 3(3):227–253.

Hinton, G. E. and Salakhutdinov, R. R. (2006). Reducing the dimensionality of data with neural networks. science, 313(5786):504–507.

Hiransha, M., Gopalakrishnan, E., Menon, V. K., and Soman, K. (2018). Nse stock market prediction using deep-learning models. Procedia Computer Science, 132:1351–1362. International Conference on Computational Intelligence and Data Science.

Kingma, D. P. and Welling, M. (2013). Auto-encoding variational bayes. arXiv preprint arXiv:1312.6114.

Krizhevsky, A., Sutskever, I., and Hinton, G. E. (2012). Imagenet classification with deep convolutional neural networks. In Pereira, F., Burges, C., Bottou, L., and Weinberger, K., editors, Advances in Neural Information Processing Systems, volume 25. Curran Associates, Inc.

Massey, F. J. (1951). The kolmogorov-smirnov test for goodness of fit. Journal of the American Statistical Association, 46(253):68–78. Milner, J. M. and Kouvelis, P. (2005). Order quantity and timing flexibility in supply chains: The role of demand characteristics. Management Science, 51(6):970–985.

Papalexopoulos, A. and Hesterberg, T. (1990). A regression-based approach to short-term system load forecasting. IEEE Transactions on Power Systems, 5(4):1535–1547.

Petruzzi, N. C. and Dada, M. (1999). Pricing and the newsvendor problem: A review with extensions. Operations research, 47(2):183–194.

Smirnov, D., Herer, Y. T., and Avrahami, A. (2021). Two-phase newsvendor with optimally timed additional replenishment: Model, algorithm, case study. Production and Operations Management, 30(9):2871–2889.

Sohn, K., Lee, H., and Yan, X. (2015). Learning structured output representation using deep conditional generative models. Advances in neural information processing systems, 28.

Villani, C. (2009). Optimal transport: old and new, volume 338. Springer. Wager, S., Hastie, T., and Efron, B. (2014). Confidence intervals for random forests: The jackknife and the infinitesimal jackknife. The journal of machine learning research, 15(1):1625–1651.

Wang, C., Sharifnia, E., Gao, Z., Tindemans, S. H., and Palensky, P. (2022). Generating multivariate load states using a conditional variational autoencoder. Electric Power Systems Research, 213:108603.

Yan, X., Yang, J., Sohn, K., and Lee, H. (2016). Attribute2image: Conditional image generation from visual attributes. In Computer Vision–ECCV 2016: 14th European Conference, Amsterdam, The Netherlands, October 11–14, 2016, Proceedings, Part IV 14, pages 776–791. Springer.