Scholarlink

Abstract <p>Milk production is a vital component of the agricultural economy in Asia; however, its variability across livestock categories poses significant challenges for forecasting and long-term planning. The absence of precise predictive tools for milk production trends hinders effective resource allocation and policy formulation, underscoring the need for advanced forecasting approaches. This study aims to develop and evaluate deep learning models for accurate milk production forecasting across buffalo, camel, cattle, goat, and sheep populations from 1961 to 2022. Three model architectures were implemented: BiLSTM, CNN, and a hybrid CNN-BiLSTM, trained with an 80/20 train–test split. Model performance was assessed using statistical indicators, including the root mean square error (RMSE) and the correlation coefficient (R). The hybrid CNN-BiLSTM demonstrated superior predictive accuracy, achieving test RMSE values of 2959.65 ‘00t (buffalo), 16.91 ‘00t (camel), and 4515.18 ‘00t (cattle), representing improvements of up to 67.7% over BiLSTM and 59.1% over CNN. Correlation coefficients were consistently high across most categories (R &gt; 0.98 for cattle, camel, and buffalo), confirming the model’s strong capability in capturing both temporal dependencies and feature patterns. These findings emphasize the potential of advanced deep learning models to enhance forecasting accuracy for agricultural production, thereby supporting informed decision-making and sustainable livestock management.</p> Graphical Abstract <p></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

Time-series analysis and forecasting of milk production across Asian livestock using BiLSTM and CNN models

  • 期刊论文

  • 发表日期：

    2026年7月20日

查看全文
  • Pradeep Mishra,
  • Bilel Zerouali,
  • Nadjem Bailek,
  • Celso Augusto Guimarães Santos,
  • Binita Kumari,
  • Yong Jie Wong,
  • Priyanka Lal,
  • Ahmed Elbeltagi

摘要

Abstract

Milk production is a vital component of the agricultural economy in Asia; however, its variability across livestock categories poses significant challenges for forecasting and long-term planning. The absence of precise predictive tools for milk production trends hinders effective resource allocation and policy formulation, underscoring the need for advanced forecasting approaches. This study aims to develop and evaluate deep learning models for accurate milk production forecasting across buffalo, camel, cattle, goat, and sheep populations from 1961 to 2022. Three model architectures were implemented: BiLSTM, CNN, and a hybrid CNN-BiLSTM, trained with an 80/20 train–test split. Model performance was assessed using statistical indicators, including the root mean square error (RMSE) and the correlation coefficient (R). The hybrid CNN-BiLSTM demonstrated superior predictive accuracy, achieving test RMSE values of 2959.65 ‘00t (buffalo), 16.91 ‘00t (camel), and 4515.18 ‘00t (cattle), representing improvements of up to 67.7% over BiLSTM and 59.1% over CNN. Correlation coefficients were consistently high across most categories (R > 0.98 for cattle, camel, and buffalo), confirming the model’s strong capability in capturing both temporal dependencies and feature patterns. These findings emphasize the potential of advanced deep learning models to enhance forecasting accuracy for agricultural production, thereby supporting informed decision-making and sustainable livestock management.

Graphical Abstract