Scholarlink

<p>Potato is one of India’s most important food and cash crops, playing a vital role in national nutrition, food security, and the agricultural economy. Accurate yield forecasting is critical for policymakers, supply chain managers, and farmers to optimize planning and mitigate market risks. This study applies long short-term memory (LSTM) neural networks to predict state-wise potato yield and production trends in India using data from 2011 to 2024 compiled from the Ministry of Agriculture and Farmers’ Welfare, FAOSTAT, and open-access databases. The objective was to evaluate the predictive capability of LSTM in comparison with traditional models such as autoregressive integrated moving average (ARIMA). The LSTM model achieved a lower root mean squared error (RMSE = 1.05) and higher coefficient of determination (<i>R</i><sup>2</sup> = 0.92) than ARIMA (RMSE = 1.78; <i>R</i><sup>2</sup> = 0.85), confirming its superiority in capturing nonlinear temporal dependencies. Forecasts indicate a moderate but steady increase in potato production with regional variations, highlighting the usefulness of deep learning for state-specific agricultural planning. The findings provide a reliable basis for integrating AI-driven forecasting into policy and supply-chain management. However, the absence of climatic and soil variables limits model generalization, suggesting future research should incorporate hybrid or remote-sensing–based frameworks for improved robustness and interpretability.</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

Long Short-term Memory Approaches for Predicting Potato Yield and Production Trends in India

  • 期刊论文

  • 发表日期：

    2026年1月19日

查看全文
  • Mohammed Ahmed Alomair,
  • Khalid Ul Islam Rather

摘要

Potato is one of India’s most important food and cash crops, playing a vital role in national nutrition, food security, and the agricultural economy. Accurate yield forecasting is critical for policymakers, supply chain managers, and farmers to optimize planning and mitigate market risks. This study applies long short-term memory (LSTM) neural networks to predict state-wise potato yield and production trends in India using data from 2011 to 2024 compiled from the Ministry of Agriculture and Farmers’ Welfare, FAOSTAT, and open-access databases. The objective was to evaluate the predictive capability of LSTM in comparison with traditional models such as autoregressive integrated moving average (ARIMA). The LSTM model achieved a lower root mean squared error (RMSE = 1.05) and higher coefficient of determination (R2 = 0.92) than ARIMA (RMSE = 1.78; R2 = 0.85), confirming its superiority in capturing nonlinear temporal dependencies. Forecasts indicate a moderate but steady increase in potato production with regional variations, highlighting the usefulness of deep learning for state-specific agricultural planning. The findings provide a reliable basis for integrating AI-driven forecasting into policy and supply-chain management. However, the absence of climatic and soil variables limits model generalization, suggesting future research should incorporate hybrid or remote-sensing–based frameworks for improved robustness and interpretability.