Scholarlink

<p>The paper explores the determinants of fiscal openness in a global context by employing the Open Budget Index and multi-dimensional machine learning tools. Using the unsupervised algorithm of K-means clustering along with several supervised algorithms such as KNN, SVM, and ANN in conjunction with XAI tools, namely SHAP and LIME, we are able to determine different regimes of governance that go beyond traditional classifications. Spatial clustering is evident, as indicated by a Moran’s I coefficient of 0.278; specifically, Western economies (Cluster 3) score the best, followed by the lowest ranked economies from the Middle East/Africa (Cluster 1). Such conclusions are corroborated by the supervised methods used in this study, with KNN providing the highest accuracy rate of 97.3%. Our SHAP analysis reveals that oversight institutions and citizen engagement are the main drivers of a state’s financial transparency, capturing the intricate, non-linear interplay between the institutional components of budget openness. Further, network analysis helps identify an “ecosystem of accountability”, characterized by a mutually reinforcing relationship between independent audits and oversight laws. Overall, the paper finds evidence of a “governance gap” in developing countries, which possess enough audit capacity, yet lack public engagement.</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

Unveiling structural patterns of global budget transparency using advanced machine learning and explainable artificial intelligence frameworks

  • 期刊论文

  • 发表日期：

    2026年6月3日

查看全文
  • Sadullah Çelik,
  • Sercan Yavan,
  • Gamze Çimen

摘要

The paper explores the determinants of fiscal openness in a global context by employing the Open Budget Index and multi-dimensional machine learning tools. Using the unsupervised algorithm of K-means clustering along with several supervised algorithms such as KNN, SVM, and ANN in conjunction with XAI tools, namely SHAP and LIME, we are able to determine different regimes of governance that go beyond traditional classifications. Spatial clustering is evident, as indicated by a Moran’s I coefficient of 0.278; specifically, Western economies (Cluster 3) score the best, followed by the lowest ranked economies from the Middle East/Africa (Cluster 1). Such conclusions are corroborated by the supervised methods used in this study, with KNN providing the highest accuracy rate of 97.3%. Our SHAP analysis reveals that oversight institutions and citizen engagement are the main drivers of a state’s financial transparency, capturing the intricate, non-linear interplay between the institutional components of budget openness. Further, network analysis helps identify an “ecosystem of accountability”, characterized by a mutually reinforcing relationship between independent audits and oversight laws. Overall, the paper finds evidence of a “governance gap” in developing countries, which possess enough audit capacity, yet lack public engagement.