Scholarlink

<p>In this study, an integrated, AI-driven framework is recommended for enhancing Management Information Systems (MIS) capabilities to fulfill the demands of studying sustainable supply chain optimization and environmental impact analysis. To address the three core tasks, research utilizes both classical machine learning regression models (Linear Regression, Random Forest, Gradient Boosting, XGBoost, SVR) as well as a deep learning-based Gated TabTransformer (GTT) to predict GHG emissions, a Variational Graph Autoencoder (VGAE) for unsupervised anomaly detection, and a hybrid interpretability model constructed with Explainable Boosting Machines (EBM) and SHapley Additive exPlanations (SHAP) for explainable AI. The results show that the Linear Regression and ensemble models (e.g., Gradient Boosting with <i>R</i><sup>2</sup> = 0.9995 and MAE = 0.0036) clearly outperform the existing methods both in terms of accuracy and robustness. In addition, the GTT shows strong generalization (<i>R</i><sup>2</sup> = 0.9883), and the VGAE is able to detect anomalies in emission patterns by learning both topological and feature-level deviations. In addition, via EBM–SHAP, the framework enables interpretable local and global insights for data-driven decision making, with transparency. In addition, the proposed framework is a comprehensive solution that bridges sustainability, AI, and MIS; it improves the state-of-the-art of sustainability analytics on the supply chain.</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

Leveraging artificial intelligence in management information systems for sustainable supply chain optimization and environmental impact analysis

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年12月24日

查看全文
  • Md Tanvir Rahman Tarafder,
  • Md Eahia Ansari,
  • Md Ariful Alam,
  • Sanjib Kumar Shil,
  • Rakibul Islam,
  • Khandakar Rabbi Ahmed

摘要

In this study, an integrated, AI-driven framework is recommended for enhancing Management Information Systems (MIS) capabilities to fulfill the demands of studying sustainable supply chain optimization and environmental impact analysis. To address the three core tasks, research utilizes both classical machine learning regression models (Linear Regression, Random Forest, Gradient Boosting, XGBoost, SVR) as well as a deep learning-based Gated TabTransformer (GTT) to predict GHG emissions, a Variational Graph Autoencoder (VGAE) for unsupervised anomaly detection, and a hybrid interpretability model constructed with Explainable Boosting Machines (EBM) and SHapley Additive exPlanations (SHAP) for explainable AI. The results show that the Linear Regression and ensemble models (e.g., Gradient Boosting with R2 = 0.9995 and MAE = 0.0036) clearly outperform the existing methods both in terms of accuracy and robustness. In addition, the GTT shows strong generalization (R2 = 0.9883), and the VGAE is able to detect anomalies in emission patterns by learning both topological and feature-level deviations. In addition, via EBM–SHAP, the framework enables interpretable local and global insights for data-driven decision making, with transparency. In addition, the proposed framework is a comprehensive solution that bridges sustainability, AI, and MIS; it improves the state-of-the-art of sustainability analytics on the supply chain.