Transforming oil market analysis: A novel GAN + LSTM predictive framework

Next Energy Pub Date : 2025-04-01 DOI:10.1016/j.nxener.2025.100303

Prity Kumari , G.Y. Chandan , Satish Kumar M

{"title":"Transforming oil market analysis: A novel GAN + LSTM predictive framework","authors":"Prity Kumari , G.Y. Chandan , Satish Kumar M","doi":"10.1016/j.nxener.2025.100303","DOIUrl":null,"url":null,"abstract":"<div><div>A novel method of predicting the crude oil WTI futures prices based on a data set covering April 12, 2009 through January 7, 2024. To capture complex market dynamics more precisely, it incorporates key market factors such as open, high, and low price along with slacked variable inclusions and moving averages over 7, 14, and 30 days. It includes generative adversarial networks augmented with long short-term memory (GAN + LSTM), LSTM, gated recurrent units (GRUs), and artificial neural networks (ANNs) as predictive models, where their performance was compared by a variety of measurements like mean squared error (MSE), mean absolute error (MAE), mean absolute percentage error (MAPE), symmetric mean absolute percentage error (SMAPE), normalized root mean square error (NRMSE), and adjusted R-square. The GAN + LSTM model proved its accuracy over others, with the lowest MSE (0.001), MAE (0.029), MAPE (4.639), SMAPE (4.734), NRMSE (0.057), and the highest adjusted R-squared (0.943). This model has been viewed as dominant due to its ability to integrate and gain information from complex data patterns. In volatile markets, the outcome of this study will be of fundamental significance in developing proactive models that will provide the most efficient means of making more informed decisions.</div></div>","PeriodicalId":100957,"journal":{"name":"Next Energy","volume":"7 ","pages":"Article 100303"},"PeriodicalIF":0.0000,"publicationDate":"2025-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Next Energy","FirstCategoryId":"1085","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S2949821X25000663","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}

引用次数: 0

Abstract

A novel method of predicting the crude oil WTI futures prices based on a data set covering April 12, 2009 through January 7, 2024. To capture complex market dynamics more precisely, it incorporates key market factors such as open, high, and low price along with slacked variable inclusions and moving averages over 7, 14, and 30 days. It includes generative adversarial networks augmented with long short-term memory (GAN + LSTM), LSTM, gated recurrent units (GRUs), and artificial neural networks (ANNs) as predictive models, where their performance was compared by a variety of measurements like mean squared error (MSE), mean absolute error (MAE), mean absolute percentage error (MAPE), symmetric mean absolute percentage error (SMAPE), normalized root mean square error (NRMSE), and adjusted R-square. The GAN + LSTM model proved its accuracy over others, with the lowest MSE (0.001), MAE (0.029), MAPE (4.639), SMAPE (4.734), NRMSE (0.057), and the highest adjusted R-squared (0.943). This model has been viewed as dominant due to its ability to integrate and gain information from complex data patterns. In volatile markets, the outcome of this study will be of fundamental significance in developing proactive models that will provide the most efficient means of making more informed decisions.

查看原文本刊更多论文

改变石油市场分析：一种新的GAN + LSTM预测框架

基于2009年4月12日至2024年1月7日的数据集，一种预测WTI原油期货价格的新方法。为了更精确地捕捉复杂的市场动态，它结合了关键的市场因素，如开盘、高价和低价，以及松散的变量内含物和7、14和30天的移动平均线。它包括具有长短期记忆增强的生成式对抗网络（GAN + LSTM）， LSTM，门控制循环单元（gru）和人工神经网络（ann）作为预测模型，其中它们的性能通过各种测量进行比较，如均方误差（MSE），平均绝对误差（MAE），平均绝对百分比误差（MAPE），对称平均绝对百分比误差（SMAPE），标准化均方根误差（NRMSE）和调整r方。GAN + LSTM模型的MSE最低（0.001），MAE最低（0.029），MAPE最低（4.639），SMAPE最低（4.734），NRMSE最低（0.057），调整后r方最高（0.943）。由于能够从复杂的数据模式中集成和获取信息，该模型已被视为占主导地位。在动荡的市场中，本研究的结果将对开发前瞻性模型具有根本意义，这些模型将提供做出更明智决策的最有效手段。

本文章由计算机程序翻译，如有差异，请以英文原文为准。

求助全文

约1分钟内获得全文求助全文

来源期刊

Next Energy

自引率

0.00%

发文量