石油化學工業，以下簡稱石化工業(petrochemical industry)是世界經濟中相當重要的產業之一，多數應用於人類食、衣、住、行、育、樂的產品都是由石化工業製造，而其中影響該產業運作最重要的原物料就是原油，因此若可以以準確預測原油價格掌握未來趨勢變化，將可以協助企業降低原油採購成本及採購風險，進而提升企業營運績效。
本研究提出兩階段的方法進行原油價格預測及原油採購決策，首先第一階段原油價格預測，因為影響原油價格的因子眾多，舉凡個體經濟學、總體經濟學、原油產量、儲量等因子，或是地緣政治等因素，都會造成原油價格的變動，因此首先會用資料探勘的技術從眾多變數中找出影響油價的關鍵因子，並利用格蘭傑因果關係檢定(Granger causality test )探討重要變數間相互影響的因果關係， 接著訓練三種不同機器學習的模型分別為支持向量迴歸(Support Vector Regression)、 長短期記憶遞歸神經網路 (Long Short-Term Memory) 梯度提升回歸樹(Gradient Boosting Regression)透過模型整合的方法，將不同模型的優勢整合起來，進行油價的預測並透過實際資料驗證預測結果 。第二階段原油採購決策則會加入歷史庫存資料，並利用第一階段的油價預測結果,透過本研究提出的條件風險價值強化學習(Conditional Value at Risk Deep Reinforcement Learning )建構採購決策模型,藉此從經驗中學習,進而產生採購策略,最終希望同時使總採購成本、庫存成本、缺貨成本及採購風險最小化,並提升企業整體效益,以達成預測輔助決策之目的。

The petrochemical industry is one of the most important industries in the world economy. Most of the products used in human food, clothing, housing, transportation, education, and entertainment are manufactured by the petrochemical industry, and the most important raw materials affecting the operation of the industry are Crude oil, therefore, if you can accurately forecast the price of crude oil to grasp future trends, it will help companies reduce crude oil procurement costs and procurement risks, thereby improving corporate operating performance.
This study proposes a two-stage method for crude oil price forecasting and crude oil purchase decisions. First, crude oil price forecasting in the first stage, because there are many factors that affect crude oil prices, such as individual economics, overall economics, crude oil production, reserves and other factors, or it is geopolitics and other factors that will cause changes in crude oil prices. Therefore, firstly, we will use data exploration technology to find the key factors affecting oil prices from many variables, and use the Granger causality test to explore the important variables. The causality of mutual influence, and then train three different machine learning models: Support Vector Regression, Long Short-Term Memory, and Gradient Boosting Regression through the model The integrated method integrates the advantages of different models to make oil price forecasts and verify the forecast results through actual data. In the second stage of crude oil procurement decisions, historical inventory data will be added, and the results of the first stage of oil price forecasting will be used to construct the procurement decision model through Conditional Value at Risk Deep Reinforcement Learning proposed in this study. Learning from experience leads to purchasing strategies, and ultimately hopes to minimize the total purchasing cost, inventory cost, out-of-stock cost and purchasing risk at the same time, and improve the overall efficiency of the enterprise, so as to achieve the purpose of forecasting and assisting decision-making.

Abramson, B., & Finizza, A. J. (1991). A belief network implementation of target capacity utilization. Paper presented at the Energy Disruptions: Lessons, Opportunities & Prospects, 13th IAEE North American Conference, 1991.
Artzner, P., Delbaen, F., Eber, J. M., & Heath, D. (1999). Coherent measures of risk. Mathematical finance, 9(3), 203-228.
Azadeh, A., Moghaddam, M., Khakzad, M., & Ebrahimipour, V. (2012). A flexible neural network-fuzzy mathematical programming algorithm for improvement of oil price estimation and forecasting. Computers & Industrial Engineering, 62(2), 421-430.
Aziz, N., Abdullah, M. H. A., & Zaidi, A. N. (2020, 8-9 Oct. 2020). Predictive Analytics for Crude Oil Price Using RNN-LSTM Neural Network. Paper presented at the 2020 International Conference on Computational Intelligence (ICCI).
Bacon, R. (1991). Modelling the price of oil. Oxford review of economic policy, 7(2), 17-34.
Bai-Jian Chou (周百建 ), C.-Y. L. (2018). Data Mining For Price Forecasting of Petrochemical Raw Material. 2018 INFORMS International Conference.
Bellemare, M. G., Dabney, W., & Munos, R. (2017). A distributional perspective on reinforcement learning. Paper presented at the International Conference on Machine Learning.
Bernhard, J., Pollok, S., & Knoll, A. (2019, 9-12 June 2019). Addressing Inherent Uncertainty: Risk-Sensitive Behavior Generation for Automated Driving using Distributional Reinforcement Learning. Paper presented at the 2019 IEEE Intelligent Vehicles Symposium (IV).
Bostanchi, H. (2018). WTI Oil Price Prediction Modeling and Forecasting.
Chinn, M. D., LeBlanc, M., & Coibion, O. (2005). The predictive content of energy futures: an update on petroleum, natural gas, heating oil and gasoline. Retrieved from
Dabney, W., Ostrovski, G., Silver, D., & Munos, R. (2018). Implicit quantile networks for distributional reinforcement learning. Paper presented at the International conference on machine learning.
Dabney, W., Rowland, M., Bellemare, M. G., & Munos, R. (2018). Distributional reinforcement learning with quantile regression. Paper presented at the Thirty-Second AAAI Conference on Artificial Intelligence.
Dagum, E. (2013). Time Series Modelling and Decomposition. Statistica, 70. doi:10.6092/issn.1973-2201/3597
Erlenkotter, D. (1990). Ford Whitman Harris and the Economic Order Quantity Model. Operations Research, 38, 937-946. doi:10.1287/opre.38.6.937
Gao, S., & Lei, Y. (2017). A new approach for crude oil price prediction based on stream learning. Geoscience Frontiers, 8(1), 183-187. Retrieved from https://www.sciencedirect.com/science/article/pii/S167498711630086X. doi:https://doi.org/10.1016/j.gsf.2016.08.002
Gong, X., & Lin, B. (2017). Forecasting the good and bad uncertainties of crude oil prices using a HAR framework. Energy Economics, 67, 315-327. Retrieved from https://www.sciencedirect.com/science/article/pii/S014098831730302X. doi:https://doi.org/10.1016/j.eneco.2017.08.035
Haidar, I., Kulkarni, S., & Pan, H. (2008). Forecasting model for crude oil prices based on artificial neural networks. Paper presented at the 2008 International Conference on Intelligent Sensors, Sensor Networks and Information Processing.
Hull, J., & White, A. (1998). Incorporating volatility updating into the historical simulation method for value-at-risk. Journal of risk, 1(1), 5-19.
Jiang, H., Li, Y., Zhou, C., Hong, H., Glade, T., & Yin, K. (2020). Landslide Displacement Prediction Combining LSTM and SVR Algorithms: A Case Study of Shengjibao Landslide from the Three Gorges Reservoir Area. Applied Sciences, 10(21), 7830.
Jinliang, Z., Mingming, T., & Mingxin, T. (2009). Effects simulation of international gold prices on crude oil prices based on WBNNK model. Paper presented at the 2009 ISECS International Colloquium on Computing, Communication, Control, and Management.
Kara, A., & Dogan, I. (2018). Reinforcement learning approaches for specifying ordering policies of perishable inventory systems. Expert Systems with Applications, 91, 150-158. Retrieved from https://www.sciencedirect.com/science/article/pii/S0957417417305900. doi:https://doi.org/10.1016/j.eswa.2017.08.046
Katanyukul, T., & Chong, E. K. P. (2014). Intelligent Inventory Control via Ruminative Reinforcement Learning. Journal of Applied Mathematics, 2014, 238357. Retrieved from https://doi.org/10.1155/2014/238357. doi:10.1155/2014/238357
Katanyukul, T., Duff, W. S., & Chong, E. K. P. (2011). Approximate dynamic programming for an inventory problem: Empirical comparison. Computers & Industrial Engineering, 60(4), 719-743. Retrieved from https://www.sciencedirect.com/science/article/pii/S0360835211000246. doi:https://doi.org/10.1016/j.cie.2011.01.007
Lee, C.-Y., Hung, Yu-Hsin, and Chen, Yen-Wen. (2021). Hybrid Data Science and Reinforcement Learning in Data Envelopment Analysis.
Lyle, C., Bellemare, M. G., & Castro, P. S. (2019). A comparative analysis of expected and distributional reinforcement learning. Paper presented at the Proceedings of the AAAI Conference on Artificial Intelligence.
Ma, R., Yang, T., Breaz, E., Li, Z., Briois, P., & Gao, F. (2018). Data-driven proton exchange membrane fuel cell degradation predication through deep learning method. Applied Energy, 231, 102-115. Retrieved from http://www.sciencedirect.com/science/article/pii/S0306261918314181. doi:https://doi.org/10.1016/j.apenergy.2018.09.111
Mihatsch, O., & Neuneier, R. (2002). Risk-Sensitive Reinforcement Learning. Machine Learning, 49(2), 267-290. Retrieved from https://doi.org/10.1023/A:1017940631555. doi:10.1023/A:1017940631555
Mirmirani, S., & Li, H. C. (2004). A comparison of VAR and neural networks with genetic algorithm in forecasting price of oil. In Applications of artificial intelligence in finance and economics: Emerald Group Publishing Limited.
Mnih, V., Kavukcuoglu, K., Silver, D., Graves, A., Antonoglou, I., Wierstra, D., & Riedmiller, M. (2013). Playing atari with deep reinforcement learning. arXiv preprint arXiv:1312.5602.
Morana, C. (2001). A semiparametric approach to short-term oil price forecasting. Energy Economics, 23(3), 325-338.
Nelson, Y., Stoner, S., & Office, C. E. C. F. R. (1996). Results of the Delphi VIII Survey of Oil Price Forecasts: Staff Report: California Energy Commission.
Panas, E., & Ninni, V. (2000). Are oil markets chaotic? A non-linear dynamic analysis. Energy Economics, 22(5), 549-568. Retrieved from https://www.sciencedirect.com/science/article/pii/S0140988300000499. doi:https://doi.org/10.1016/S0140-9883(00)00049-9
Perez, H. D., Hubbs, C. D., Li, C., & Grossmann, I. E. (2021). Algorithmic Approaches to Inventory Management Optimization. Processes, 9(1), 102. Retrieved from https://www.mdpi.com/2227-9717/9/1/102.
Rockafellar, R. T., & Uryasev, S. (2000). Optimization of conditional value-at-risk. Journal of risk, 2, 21-42.
Safari, A., & Davallou, M. (2018). Oil price forecasting using a hybrid model. Energy, 148, 49-58.
Wu, Y., Yuan, M., Dong, S., Lin, L., & Liu, Y. (2018). Remaining useful life estimation of engineered systems using vanilla LSTM neural networks. Neurocomputing, 275, 167-179. Retrieved from http://www.sciencedirect.com/science/article/pii/S0925231217309505. doi:https://doi.org/10.1016/j.neucom.2017.05.063
Xie, W., Yu, L., Xu, S., & Wang, S. (2006a). A new method for crude oil price forecasting based on support vector machines. Paper presented at the International conference on computational science.
Xie, W., Yu, L., Xu, S., & Wang, S. (2006b, 2006//). A New Method for Crude Oil Price Forecasting Based on Support Vector Machines. Paper presented at the Computational Science – ICCS 2006, Berlin, Heidelberg.
Zhang, X., Lai, K. K., & Wang, S.-Y. (2008). A new approach for crude oil price analysis based on empirical mode decomposition. Energy Economics, 30(3), 905-918.
Zhou, J., Zhang, S., & Li, Y. (2018, 8-9 Dec. 2018). A Deep Q-Learning Approach for Continuous Review Policies with Uncertain Lead Time Demand Patterns. Paper presented at the 2018 11th International Symposium on Computational Intelligence and Design (ISCID).