Scholarlink

<p>Agriculture and agri-food systems contribute significantly to climate change through greenhouse gas emissions that are pushing global temperatures higher by more than 1.5&#xa0;°C. It is essential to identify the key factors contributing to rising temperatures and to implement measures to mitigate them to prevent further deterioration in the future. This paper introduced an approach called (XAI-SHAP) that utilizes deep learning models in cooperation with an Explainable AI method (XAI) called Shapley Additive Explanations (SHAP) to identify agricultural activities that are significantly responsible for global temperature rise. This work depends on using an agri-food CO2 emission dataset to investigate the relationship between agri-food sector CO2 emissions and temperature rise. This proposed approach, XAI-SHAP can predict global temperature rise and identify agricultural activities that are major contributors to it. The XAI-SHAP includes five main phases: data preprocessing phase, optimization, training phase, and testing phase, and finally explanation phase. The proposed XAI-SHAP accurately predicts the rate of global temperature rise using a Deep Neural Network (DNN) model whose hyperparameters are determined by a swarm optimization algorithm called the Moth Search Algorithm (MSA). The SHAP method is applied in the last phase to offer a clear interpretation of model predictions by highlighting features with positive and negative impacts on XAI-SHAP outputs. The proposed XAI-SHAP can predict the average temperature rise rate with high efficiency, as it achieved an average test mean squared error (MSE) of 0.0147 and a mean absolute error (MAE) of 0.0832 across five-fold cross-validation, with superior inference efficiency of 0.0003&#xa0;s compared to competing models. XAI-SHAP approach results show that crop residues, drained organic soils, agri-food waste disposal, pesticide manufacturing, and food packaging are the most significant agricultural activities contributing to global warming. These findings highlight the need for prioritized mitigation actions targeting agricultural activities that significantly contribute to greenhouse gas emissions.</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

An optimized and explainable artificial intelligence approach for environmental impact of the agri-food industry

  • 期刊论文

  • 发表日期：

    2026年3月14日

查看全文
  • Dalia Ezzat,
  • Mona M. Soliman,
  • Ashraf Darwish,
  • Aboul Ella Hassnien

摘要

Agriculture and agri-food systems contribute significantly to climate change through greenhouse gas emissions that are pushing global temperatures higher by more than 1.5 °C. It is essential to identify the key factors contributing to rising temperatures and to implement measures to mitigate them to prevent further deterioration in the future. This paper introduced an approach called (XAI-SHAP) that utilizes deep learning models in cooperation with an Explainable AI method (XAI) called Shapley Additive Explanations (SHAP) to identify agricultural activities that are significantly responsible for global temperature rise. This work depends on using an agri-food CO2 emission dataset to investigate the relationship between agri-food sector CO2 emissions and temperature rise. This proposed approach, XAI-SHAP can predict global temperature rise and identify agricultural activities that are major contributors to it. The XAI-SHAP includes five main phases: data preprocessing phase, optimization, training phase, and testing phase, and finally explanation phase. The proposed XAI-SHAP accurately predicts the rate of global temperature rise using a Deep Neural Network (DNN) model whose hyperparameters are determined by a swarm optimization algorithm called the Moth Search Algorithm (MSA). The SHAP method is applied in the last phase to offer a clear interpretation of model predictions by highlighting features with positive and negative impacts on XAI-SHAP outputs. The proposed XAI-SHAP can predict the average temperature rise rate with high efficiency, as it achieved an average test mean squared error (MSE) of 0.0147 and a mean absolute error (MAE) of 0.0832 across five-fold cross-validation, with superior inference efficiency of 0.0003 s compared to competing models. XAI-SHAP approach results show that crop residues, drained organic soils, agri-food waste disposal, pesticide manufacturing, and food packaging are the most significant agricultural activities contributing to global warming. These findings highlight the need for prioritized mitigation actions targeting agricultural activities that significantly contribute to greenhouse gas emissions.