Scholarlink

<p>Supply chain risk management has emerged as a critical area of focus, as disruptions in this area can significantly impact the overall performance and profitability of organizations. In today’s dynamic and complex global environment, Supply Chains confront escalating challenges, such as demand volatility, globalization, disruptions, and complexity. Traditional approaches to Supply Chain Management cannot predict risks accurately due to their computational complexity and overfitting issues. The proposed framework overcomes the aforementioned challenges in the existing techniques by introducing the Elite attentive foraging optimized incremental distributed learning-based deep Gradient Boosting (EAF-optimized IDL-based Deep GB) model to advance the supply chain risk prediction. Utilizing the K-Nearest Neighbor imputation method, the research addresses missing data, enhancing dataset completeness and reliability. The Synthetic Minority Over-Sampling Technique is utilized to minimize the risk of overfitting and an incremental distributed learning technique mitigates model bias and bolsters predictive capabilities, respectively. Additionally, a Convolutional Neural Network is employed to extract relevant features from supply chain data, optimizing pattern detection in complex datasets. The model leverages a Light Gradient Boosting Machine and an ensemble approach to improve predictive accuracy by offering a nuanced understanding of supply chain data features. Specifically, the EAF optimization fine-tunes the hyperparameters of the IDL-based deep GB model to improve the prediction accuracy. Owing to the reliable risk management system, the performance evaluation of the EAF-optimized IDL-based deep GB model demonstrates notable effectiveness with 90% of training achieving 98% accuracy, 98% F-measure, 97.69% precision, 98.31% recall, 98.31% sensitivity, and 98.24% specificity outperforming other state-of-the-art methods.</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

Supply chain risk prediction using elite attentive foraging optimized incremental distributed learning based deep gradient boosting model

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年6月5日

查看全文
  • Asmaa Kafou,
  • Ahmad Alzubi,
  • Tolga Oz

摘要

Supply chain risk management has emerged as a critical area of focus, as disruptions in this area can significantly impact the overall performance and profitability of organizations. In today’s dynamic and complex global environment, Supply Chains confront escalating challenges, such as demand volatility, globalization, disruptions, and complexity. Traditional approaches to Supply Chain Management cannot predict risks accurately due to their computational complexity and overfitting issues. The proposed framework overcomes the aforementioned challenges in the existing techniques by introducing the Elite attentive foraging optimized incremental distributed learning-based deep Gradient Boosting (EAF-optimized IDL-based Deep GB) model to advance the supply chain risk prediction. Utilizing the K-Nearest Neighbor imputation method, the research addresses missing data, enhancing dataset completeness and reliability. The Synthetic Minority Over-Sampling Technique is utilized to minimize the risk of overfitting and an incremental distributed learning technique mitigates model bias and bolsters predictive capabilities, respectively. Additionally, a Convolutional Neural Network is employed to extract relevant features from supply chain data, optimizing pattern detection in complex datasets. The model leverages a Light Gradient Boosting Machine and an ensemble approach to improve predictive accuracy by offering a nuanced understanding of supply chain data features. Specifically, the EAF optimization fine-tunes the hyperparameters of the IDL-based deep GB model to improve the prediction accuracy. Owing to the reliable risk management system, the performance evaluation of the EAF-optimized IDL-based deep GB model demonstrates notable effectiveness with 90% of training achieving 98% accuracy, 98% F-measure, 97.69% precision, 98.31% recall, 98.31% sensitivity, and 98.24% specificity outperforming other state-of-the-art methods.