Scholarlink

<p>Wheat is a crucial staple crop for ensuring food security in India, yet improving its productivity remains essential to meet the demands of a growing population. This study aims to optimize nitrogen (N) and zinc (Zn) fertilization strategies, specifically by reducing N inputs through nano-urea foliar applications, while leveraging machine learning (ML) approaches to improve the accuracy of wheat yield prediction. A field dataset was collected during the <i>rabi</i> seasons of 2021–22 and 2022–23. The study assessed the integration of key yield attributes, nutrient concentrations, and nutrient uptake into ML-based prediction models. Six ML algorithms were employed: k-Nearest Neighbors (kNN), Random Forest (RF), Support Vector Regression (SVR), Multivariate Adaptive Regression Splines (MARS), Gradient Boosting Machine (GBM), and an ensemble model (GBM + SVR). Model performance was assessed using the Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and R<sup>2</sup> values. Hyperparameter tuning was performed to enhance model accuracy. Among the models tested, MARS and RF delivered the best performance. The MARS model achieved the highest prediction accuracy with MAE values of 0.025 and 0.007 t ha<sup>− 1</sup>, RMSE values of 0.013 and 0.009 t ha<sup>− 1</sup>, and R<sup>2</sup> = 0.99 for both years. These models also revealed the most influential factors affecting wheat yield, with ML-based models, particularly MARS and RF, demonstrating strong potential in predicting wheat yield and optimizing N and Zn fertilization. These models provide valuable insights for improving sustainable intensification and optimizing nutrient management in wheat cultivation.</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

Integrating experimental data and machine learning models for wheat yield estimation

  • 期刊论文

  • 发表日期：

    2026年4月7日

查看全文
  • Kadapa Sreenivasa Reddy,
  • Yashbir Singh Shivay,
  • Radha Prasanna,
  • Dinesh Kumar,
  • Panneerselvam Peramaiyan,
  • Suryakanta Kashyap,
  • Sunil Mandi,
  • Somanath Nayak,
  • Kirttiranjan Baral,
  • Rohit Bapurao Borate

摘要

Wheat is a crucial staple crop for ensuring food security in India, yet improving its productivity remains essential to meet the demands of a growing population. This study aims to optimize nitrogen (N) and zinc (Zn) fertilization strategies, specifically by reducing N inputs through nano-urea foliar applications, while leveraging machine learning (ML) approaches to improve the accuracy of wheat yield prediction. A field dataset was collected during the rabi seasons of 2021–22 and 2022–23. The study assessed the integration of key yield attributes, nutrient concentrations, and nutrient uptake into ML-based prediction models. Six ML algorithms were employed: k-Nearest Neighbors (kNN), Random Forest (RF), Support Vector Regression (SVR), Multivariate Adaptive Regression Splines (MARS), Gradient Boosting Machine (GBM), and an ensemble model (GBM + SVR). Model performance was assessed using the Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and R2 values. Hyperparameter tuning was performed to enhance model accuracy. Among the models tested, MARS and RF delivered the best performance. The MARS model achieved the highest prediction accuracy with MAE values of 0.025 and 0.007 t ha− 1, RMSE values of 0.013 and 0.009 t ha− 1, and R2 = 0.99 for both years. These models also revealed the most influential factors affecting wheat yield, with ML-based models, particularly MARS and RF, demonstrating strong potential in predicting wheat yield and optimizing N and Zn fertilization. These models provide valuable insights for improving sustainable intensification and optimizing nutrient management in wheat cultivation.