Scholarlink

<p>This investigation evaluated the effectiveness of several machine learning models, such as LSTM, GRU, BiLSTM, SVM, and random forest, in forecasting sugarcane production in different regions of Africa and globally. The results show that geographical characteristics and production parameters greatly affect model efficacy. LSTM and GRU capture temporal dependencies better in Africa, Southern Africa, and sub-Saharan Africa with dynamic and changeable production trends, resulting in lower error metrics and accurate forecasts. The support vector machine (SVM) is the most effective model in steady producing zones, having less variability in production like Central Africa and West Africa and in global forecasting due to its scalability and simplicity. Random forest struggled with time series data, delivering disappointing results across all regions. The 2030 projections are useful in agricultural planning and resource management. African production is likely to rise steadily, and it is expected that Middle Africa will witness major expansion in sugarcane production after 2025. Sub-Saharan Africa is expected to stabilize, but global sugarcane output may drop, suggesting agricultural changes. The forecasted values from these models indicate that sugarcane production in Africa is expected to reach 110,845&#xa0;Mg by 2030, demonstrating steady growth. The projections emphasis the need for improved machine learning algorithms in agriculture's long-term planning and decision-making. Despite these promising results, more research is needed to improve forecast accuracy and relevance.</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

State-of-the-Art Sugarcane Production in African Region

  • 期刊论文

  • 发表日期：

    2025年8月7日

查看全文
  • Pradeep Mishra,
  • Mostafa Abotaleb,
  • Binita Kumari,
  • El-Sayed M. El-kenawy,
  • Shikha Yadav

摘要

This investigation evaluated the effectiveness of several machine learning models, such as LSTM, GRU, BiLSTM, SVM, and random forest, in forecasting sugarcane production in different regions of Africa and globally. The results show that geographical characteristics and production parameters greatly affect model efficacy. LSTM and GRU capture temporal dependencies better in Africa, Southern Africa, and sub-Saharan Africa with dynamic and changeable production trends, resulting in lower error metrics and accurate forecasts. The support vector machine (SVM) is the most effective model in steady producing zones, having less variability in production like Central Africa and West Africa and in global forecasting due to its scalability and simplicity. Random forest struggled with time series data, delivering disappointing results across all regions. The 2030 projections are useful in agricultural planning and resource management. African production is likely to rise steadily, and it is expected that Middle Africa will witness major expansion in sugarcane production after 2025. Sub-Saharan Africa is expected to stabilize, but global sugarcane output may drop, suggesting agricultural changes. The forecasted values from these models indicate that sugarcane production in Africa is expected to reach 110,845 Mg by 2030, demonstrating steady growth. The projections emphasis the need for improved machine learning algorithms in agriculture's long-term planning and decision-making. Despite these promising results, more research is needed to improve forecast accuracy and relevance.