Scholarlink

<p>Accurate forecasting of agricultural yields is essential for food security, climate risk management, and sustainable crop planning. Traditional regression models often struggle to capture nonlinear relationships between climatic variables and crop performance, leading to persistent prediction errors. This study develops forward-looking forecasting models, rather than retrospective regression, by training on historical IMD and DCCD datasets (1966–2023) and validating predictions on unseen future partitions. Hybrid artificial intelligence models were introduced, including Multiple Linear Regression (MLR) combined with Backpropagation Neural Network (BPNN), and XGBoost integrated with BPNN, optimized using Genetic Algorithms (GA) to improve forecasting precision. The models were trained on meteorological data from the India Meteorological Department (IMD) and cashew production records from the Directorate of Cashewnut &amp; Cocoa Development (DCCD), covering major cashew-growing regions in India. Comparative analysis shows that hybrid models outperform classical regression and standalone neural networks, with GA optimization improving R² from 0.94 to 0.978. SHAP analysis identified total cultivated area as the most influential predictor. These findings support the use of AI-enhanced forecasting tools for climate-resilient agriculture and strategic decision-making.</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

AI-based hybrid forecasting of cashew yield using multi-model techniques

  • 期刊论文

  • 发表日期：

    2026年4月17日

查看全文
  • Guhan Velusamy,
  • Dharma Raju Akasapu,
  • Nagaratna Kopparthi,
  • Sheshakumar Goroshi

摘要

Accurate forecasting of agricultural yields is essential for food security, climate risk management, and sustainable crop planning. Traditional regression models often struggle to capture nonlinear relationships between climatic variables and crop performance, leading to persistent prediction errors. This study develops forward-looking forecasting models, rather than retrospective regression, by training on historical IMD and DCCD datasets (1966–2023) and validating predictions on unseen future partitions. Hybrid artificial intelligence models were introduced, including Multiple Linear Regression (MLR) combined with Backpropagation Neural Network (BPNN), and XGBoost integrated with BPNN, optimized using Genetic Algorithms (GA) to improve forecasting precision. The models were trained on meteorological data from the India Meteorological Department (IMD) and cashew production records from the Directorate of Cashewnut & Cocoa Development (DCCD), covering major cashew-growing regions in India. Comparative analysis shows that hybrid models outperform classical regression and standalone neural networks, with GA optimization improving R² from 0.94 to 0.978. SHAP analysis identified total cultivated area as the most influential predictor. These findings support the use of AI-enhanced forecasting tools for climate-resilient agriculture and strategic decision-making.