Scholarlink

<p>This study proposes a novel multi-stage artificial intelligence (AI) methodology to predict the key drivers of integrated reporting (IR) quality (IRQ). Regarding the violation of the normality assumption in traditional methods, this study integrates clustering (K-means++), prediction (random forest, decision tree, extreme gradient boosting), and model explanation (Shapley additive explanations) techniques of AI in a consecutive way. It brings a methodological novelty, while providing more consistent results over traditional methods. Using a sample of 260 integrated reports published in 2019 and considering each pillar of the IR framework, namely, Fundamental Concepts (FC), Guiding Principles (GP), and Content Elements (CE), the model achieves strong predictive performance, with Random Forest Regressor explaining 67% of IRQ variation. The algorithms perform satisfactorily in explaining variations, showing that AI-based methods are effective on small datasets. Key predictors of IRQ involve board characteristics, financial performance indicators, and firm characteristics. The findings indicate that total assets size, GP pillar, and the number of independent members on the board have the greatest impact on IRQ. This paper also identifies best-practice clusters, providing actionable insights for practitioners and regulators. The results contribute methodologically to the AI application in accounting literature while improving IRQ.</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

Explainable artificial intelligence for determining integrated reporting quality prediction factors

  • 期刊论文

  • 发表日期：

    2026年5月7日

查看全文
  • Mert Sarioglu,
  • Gorkem Sariyer,
  • Vladimir Simic,
  • Karahan Kara

摘要

This study proposes a novel multi-stage artificial intelligence (AI) methodology to predict the key drivers of integrated reporting (IR) quality (IRQ). Regarding the violation of the normality assumption in traditional methods, this study integrates clustering (K-means++), prediction (random forest, decision tree, extreme gradient boosting), and model explanation (Shapley additive explanations) techniques of AI in a consecutive way. It brings a methodological novelty, while providing more consistent results over traditional methods. Using a sample of 260 integrated reports published in 2019 and considering each pillar of the IR framework, namely, Fundamental Concepts (FC), Guiding Principles (GP), and Content Elements (CE), the model achieves strong predictive performance, with Random Forest Regressor explaining 67% of IRQ variation. The algorithms perform satisfactorily in explaining variations, showing that AI-based methods are effective on small datasets. Key predictors of IRQ involve board characteristics, financial performance indicators, and firm characteristics. The findings indicate that total assets size, GP pillar, and the number of independent members on the board have the greatest impact on IRQ. This paper also identifies best-practice clusters, providing actionable insights for practitioners and regulators. The results contribute methodologically to the AI application in accounting literature while improving IRQ.