Scholarlink

<p>Optimizing crop yield is challenging due to complex environmental and spatial factors. Accurate yield prediction is essential for better agricultural planning, land management, and targeted interventions. The study aims to develop and evaluate a novel hybrid deep learning model, CashPred, to accurately predict cashew crop yields by leveraging advanced graph-based techniques and remote sensing imagery. CashPred combines Convolutional Neural Networks (CNN), GraphSAGE, and Dynamic Graph Neural Networks (DGNN) to integrate multi-level features. The model captures intricate spatial and temporal dynamics from remote sensing data, while feature importance analysis using SHAP values identifies the impact of key variables on crop productivity. CashPred achieved a Root Mean Squared Error (RMSE) of 0.382, a coefficient of determination (R²) of 0.956, a Mean Squared Error (MSE) of 0.287, and a Mean Absolute Error (MAE) of 0.344. The model outperformed traditional benchmarks and demonstrated strong generalization across municipalities, with Wenchi exhibiting the highest prediction accuracy. Graph visualizations and heatmap analysis highlighted critical spatial and temporal factors affecting crop productivity. The CashPred model significantly enhances cashew yield predictions by integrating spatial, temporal, and environmental factors, providing essential insights for effective agricultural planning. CashPred demonstrates the transformative potential of advanced AI in agriculture, offering a robust tool for improving yield predictions, informing policy, and promoting sustainable land management in Ghana’s high-yield cashew regions.</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

Harnessing AI for sustainable agriculture: predicting cashew yields in Ghana with cashpred

  • 期刊论文

  • 发表日期：

    2026年2月24日

查看全文
  • Bright Bediako-Kyeremeh,
  • Tinghuai Ma,
  • Emmanuel Yeboah,
  • Ben Beklisi Kwame Ayawli,
  • Jia Li,
  • Benjamin Kwapong Osibo,
  • Lorenzo Mamelona,
  • Benjamin Agyemang Badu,
  • Daniel Kumah Appiah

摘要

Optimizing crop yield is challenging due to complex environmental and spatial factors. Accurate yield prediction is essential for better agricultural planning, land management, and targeted interventions. The study aims to develop and evaluate a novel hybrid deep learning model, CashPred, to accurately predict cashew crop yields by leveraging advanced graph-based techniques and remote sensing imagery. CashPred combines Convolutional Neural Networks (CNN), GraphSAGE, and Dynamic Graph Neural Networks (DGNN) to integrate multi-level features. The model captures intricate spatial and temporal dynamics from remote sensing data, while feature importance analysis using SHAP values identifies the impact of key variables on crop productivity. CashPred achieved a Root Mean Squared Error (RMSE) of 0.382, a coefficient of determination (R²) of 0.956, a Mean Squared Error (MSE) of 0.287, and a Mean Absolute Error (MAE) of 0.344. The model outperformed traditional benchmarks and demonstrated strong generalization across municipalities, with Wenchi exhibiting the highest prediction accuracy. Graph visualizations and heatmap analysis highlighted critical spatial and temporal factors affecting crop productivity. The CashPred model significantly enhances cashew yield predictions by integrating spatial, temporal, and environmental factors, providing essential insights for effective agricultural planning. CashPred demonstrates the transformative potential of advanced AI in agriculture, offering a robust tool for improving yield predictions, informing policy, and promoting sustainable land management in Ghana’s high-yield cashew regions.