Scholarlink

<p>Over the past century, natural gas has continued to be one of the most popular energy sources and has been essential to sustaining long-term economic growth. It is a vital commodity in the world energy market, and price changes have a big impact on investment strategy, economic planning, and energy policy. Therefore, it is essential to forecast natural gas prices accurately to construct safe, efficient and low-carbon energy systems. The present study proposes a novel hybrid forecasting framework that mixes a bidirectional long short-term memory (BiLSTM) network with the Barnacles mating optimizer (BMO). Historical OHLC (open, high, low, close) price and volume data from October 2013 to September 2023 are used in the model. By optimizing the BiLSTM’s hyperparameters, the BMO improves the model’s capacity to capture temporal dependencies while reducing overfitting. To assess the model’s out-of-sample performance over a variety of time horizons and replicate actual forecasting circumstances, a rolling window forecasting technique was also used. With an <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(R^{2}\)</EquationSource> <EquationSource Format="MATHML"><math> <msup> <mi>R</mi> <mn>2</mn> </msup> </math></EquationSource> </InlineEquation> of 0.989 in the test set, the suggested BMO-BiLSTM model performs well in forecasting. A 5-fold cross-validation was performed to guarantee robustness and dependability, and the model’s capacity for generalization was examined across five more significant indices: the Nasdaq, the CSI 300, the S&amp;P 500, gold, and crude oil. The findings validate the suggested framework’s applicability beyond natural gas forecasting by showing that it maintains strong predictive accuracy across a variety of financial markets. This thorough analysis demonstrates the BMO-BiLSTM model’s potential as a potent and broadly applicable instrument for time-series forecasting in the financial and energy sectors.</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

Analyzing natural gas price projections with innovative hybrid modeling

  • 期刊论文

  • 发表日期：

    2026年3月30日

查看全文
  • Zhuo Zhao

摘要

Over the past century, natural gas has continued to be one of the most popular energy sources and has been essential to sustaining long-term economic growth. It is a vital commodity in the world energy market, and price changes have a big impact on investment strategy, economic planning, and energy policy. Therefore, it is essential to forecast natural gas prices accurately to construct safe, efficient and low-carbon energy systems. The present study proposes a novel hybrid forecasting framework that mixes a bidirectional long short-term memory (BiLSTM) network with the Barnacles mating optimizer (BMO). Historical OHLC (open, high, low, close) price and volume data from October 2013 to September 2023 are used in the model. By optimizing the BiLSTM’s hyperparameters, the BMO improves the model’s capacity to capture temporal dependencies while reducing overfitting. To assess the model’s out-of-sample performance over a variety of time horizons and replicate actual forecasting circumstances, a rolling window forecasting technique was also used. With an \(R^{2}\) R 2 of 0.989 in the test set, the suggested BMO-BiLSTM model performs well in forecasting. A 5-fold cross-validation was performed to guarantee robustness and dependability, and the model’s capacity for generalization was examined across five more significant indices: the Nasdaq, the CSI 300, the S&P 500, gold, and crude oil. The findings validate the suggested framework’s applicability beyond natural gas forecasting by showing that it maintains strong predictive accuracy across a variety of financial markets. This thorough analysis demonstrates the BMO-BiLSTM model’s potential as a potent and broadly applicable instrument for time-series forecasting in the financial and energy sectors.