Scholarlink

<p>Crude oil is an essential commodity for the entire world, and price changes have a big impact on the economy. Price volatility-related risks can be reduced for stakeholders using probabilistic crude oil forecasting. In this work, optimized deep learning (DL) models, namely convolutional neural network (CNN), long short-term memory (LSTM), gated recurrent unit (GRU), and bidirectional LSTM (BiLSTM) are used to generate prediction intervals (PI) for crude oil prices (COP) using the lower upper bound estimation (LUBE) method. Using a proprietary loss function, the LUBE method optimizes the trade-off between prediction interval normalized average width (PINAW) and prediction interval coverage probability (PICP). To evaluate the true potential of optimized DL models in probabilistic forecasting of COP using the LUBE method, monthly, weekly, and daily COP are considered, and probabilistic forecasts are obtained at different confidence levels. Four probabilistic forecast accuracy measures, namely, PICP, PINAW, average coverage error (ACE), and accumulated width deviation (AWD), are used to answer five research questions relating to COP forecasting. Additionally, the Friedman and Nemenyi hypothesis test is used to draw reliable conclusions. Simulation results suggest the statistical superiority of the optimized LSTM model in reliable probabilistic forecasting of daily, weekly, and monthly COP using the LUBE method. The optimized LSTM model with the LUBE method provided almost 100% in PICP and the lowest PINAW than all other methods considered in this study in the probabilistic forecasting of monthly, weekly, and daily COP.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A Study on Forecasting of Crude Oil Prices Employing Optimized Deep Learning Models and LUBE Method

  • 期刊论文

  • 发表日期：

    2025年7月29日

查看全文
  • Gollu Yaswanth,
  • Hanumanthu Lohith,
  • Mamuduri Jerusha,
  • Sibarama Panigrahi

摘要

Crude oil is an essential commodity for the entire world, and price changes have a big impact on the economy. Price volatility-related risks can be reduced for stakeholders using probabilistic crude oil forecasting. In this work, optimized deep learning (DL) models, namely convolutional neural network (CNN), long short-term memory (LSTM), gated recurrent unit (GRU), and bidirectional LSTM (BiLSTM) are used to generate prediction intervals (PI) for crude oil prices (COP) using the lower upper bound estimation (LUBE) method. Using a proprietary loss function, the LUBE method optimizes the trade-off between prediction interval normalized average width (PINAW) and prediction interval coverage probability (PICP). To evaluate the true potential of optimized DL models in probabilistic forecasting of COP using the LUBE method, monthly, weekly, and daily COP are considered, and probabilistic forecasts are obtained at different confidence levels. Four probabilistic forecast accuracy measures, namely, PICP, PINAW, average coverage error (ACE), and accumulated width deviation (AWD), are used to answer five research questions relating to COP forecasting. Additionally, the Friedman and Nemenyi hypothesis test is used to draw reliable conclusions. Simulation results suggest the statistical superiority of the optimized LSTM model in reliable probabilistic forecasting of daily, weekly, and monthly COP using the LUBE method. The optimized LSTM model with the LUBE method provided almost 100% in PICP and the lowest PINAW than all other methods considered in this study in the probabilistic forecasting of monthly, weekly, and daily COP.