Scholarlink

<p>Gold price forecasting remains challenging due to gold’s dual role as a financial asset and economic indicator. Traditional econometric models struggle with the asset’s nonlinear behavior, prompting increased application of machine learning approaches. This study presents a bibliometric analysis of machine learning based gold price forecasting research from 2016 to March 2026. Following PRISMA 2020 guidelines, 270 peer reviewed documents from the Web of Science Core Collection were analyzed using Bibliometrix in R and VOSviewer. The analysis mapped publication trends, geographic contributions, and intellectual structures through keyword co-occurrence, co-citation networks, and thematic mapping. Results show 8.84% annual publication growth, peaking at 47 articles in 2023. China dominates with 59.3% of publications. Hybrid deep learning models, particularly VMD-LSTM and CNN-LSTM, represent the dominant methodological approach. While gold price time series forecasting remains the central theme, transformer-based architectures are emerging. Key research gaps include limited cross market validation, insufficient uncertainty quantification, and minimal integration with related asset classes. These findings provide a roadmap for advancing gold price forecasting through more sophisticated, validated machine learning approaches.</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

Machine learning-based gold price forecasting: a bibliometric review of trends, methods, and future directions

  • 期刊论文

  • 发表日期：

    2026年7月30日

查看全文
  • Mohd Redzuan Ahmad,
  • Mohd Herwan Sulaiman

摘要

Gold price forecasting remains challenging due to gold’s dual role as a financial asset and economic indicator. Traditional econometric models struggle with the asset’s nonlinear behavior, prompting increased application of machine learning approaches. This study presents a bibliometric analysis of machine learning based gold price forecasting research from 2016 to March 2026. Following PRISMA 2020 guidelines, 270 peer reviewed documents from the Web of Science Core Collection were analyzed using Bibliometrix in R and VOSviewer. The analysis mapped publication trends, geographic contributions, and intellectual structures through keyword co-occurrence, co-citation networks, and thematic mapping. Results show 8.84% annual publication growth, peaking at 47 articles in 2023. China dominates with 59.3% of publications. Hybrid deep learning models, particularly VMD-LSTM and CNN-LSTM, represent the dominant methodological approach. While gold price time series forecasting remains the central theme, transformer-based architectures are emerging. Key research gaps include limited cross market validation, insufficient uncertainty quantification, and minimal integration with related asset classes. These findings provide a roadmap for advancing gold price forecasting through more sophisticated, validated machine learning approaches.