Scholarlink

<p>Financial decision-making increasingly requires navigating model uncertainty, estimation risk, and structural instability. Although forecasting and portfolio theory have evolved largely in parallel, both fields have developed complementary strategies for managing uncertainty through techniques such as shrinkage, model averaging, and robust optimisation. This study examines their methodological convergence through a bibliometric and thematic analysis of 489 peer-reviewed publications published between 2000 and 2025. A hybrid framework is introduced that integrates co-occurrence analysis, bibliographic coupling, fractional authorship attribution, and machine learning-based semantic clustering. Comparing a classical pipeline employing principal component analysis and K-Means with a nonlinear alternative based on uniform manifold approximation and projection combined with hierarchical density-based clustering enables an evaluation of trade-offs in cluster quality, stability, and interpretability. The analysis identifies six dominant conceptual clusters, including adaptive weighting, regularisation, and ensemble strategies, spanning both domains and reflecting a shared orientation towards robustness and ambiguity-tolerant decision-making. The research concludes by outlining opportunities for cross-disciplinary integration and proposing targeted research directions to enhance robustness, adaptability, and interpretability in forecasting and financial decision-making.</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

Forecasting meets Portfolio Theory: A Bibliometric Approach to Decision-Making under Uncertainty

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年4月17日

查看全文
  • Bernardo Raimundo,
  • Jorge Miguel Bravo

摘要

Financial decision-making increasingly requires navigating model uncertainty, estimation risk, and structural instability. Although forecasting and portfolio theory have evolved largely in parallel, both fields have developed complementary strategies for managing uncertainty through techniques such as shrinkage, model averaging, and robust optimisation. This study examines their methodological convergence through a bibliometric and thematic analysis of 489 peer-reviewed publications published between 2000 and 2025. A hybrid framework is introduced that integrates co-occurrence analysis, bibliographic coupling, fractional authorship attribution, and machine learning-based semantic clustering. Comparing a classical pipeline employing principal component analysis and K-Means with a nonlinear alternative based on uniform manifold approximation and projection combined with hierarchical density-based clustering enables an evaluation of trade-offs in cluster quality, stability, and interpretability. The analysis identifies six dominant conceptual clusters, including adaptive weighting, regularisation, and ensemble strategies, spanning both domains and reflecting a shared orientation towards robustness and ambiguity-tolerant decision-making. The research concludes by outlining opportunities for cross-disciplinary integration and proposing targeted research directions to enhance robustness, adaptability, and interpretability in forecasting and financial decision-making.