Scholarlink

<p>Traditional methods for estimating water footprints for rice production are often time-consuming and resource-intensive, highlighting the need for efficient and accurate predictive models. This study addresses this gap by evaluating the performance of seven machine learning models—Linear Regression (LR), M5P, Multi-layer Perceptron (MLP), Sequential Minimal Optimization – Support Vector Machine (SMO-SVM), Random SubSpace (RSS), Random Forest (RF), and Random Tree (RT)—in predicting the green and blue water footprints of rice in Punjab, India. Best subset regression and correlation matrix indicate that humidity, wind speed, sunshine hours, solar radiation, and total rainfall are optimal inputs for green water footprint prediction, while maximum temperature, humidity, wind speed, sunshine hours, and solar radiation are best for blue water footprint prediction. The RT model outperformed others in that, for green water footprint prediction, it achieved a correlation coefficient (CC) of 0.9991, mean absolute error (MAE) of 0.1314, root mean square error (RMSE) of 0.4553, relative absolute error (RAE) of 0.0477, and root relative squared error (RRSE) of 0.1283 during the training stage. However, during the testing stage, the RF model performed better (CC = 0.79, MAE = 154.2732, RMSE = 192.3973, RAE = 55.5602, and RRSE = 58.6433). For blue water footprint prediction, the RT model remained the best performer in both stages (training: CC = 0.9991; testing: CC = 0.9981, MAE = 0.7920, RMSE = 0.8583, RAE = 0.4440, and RRSE = 0.8290). These results suggest that machine learning can effectively support water management strategies by providing quick and reliable estimates of water footprints, which is crucial for sustainable rice production. By utilizing these models, policymakers can make informed decisions to optimize water usage and ensure sustainable agricultural practices.</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

Assessment of machine learning models to forecast water footprints of rice production

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年3月1日

查看全文
  • Ahmed Elbeltagi,
  • Aman Srivastava,
  • Durba Kashyap,
  • Leena Khadke,
  • Dinesh Kumar Vishwakarma,
  • Tripti Agarwal

摘要

Traditional methods for estimating water footprints for rice production are often time-consuming and resource-intensive, highlighting the need for efficient and accurate predictive models. This study addresses this gap by evaluating the performance of seven machine learning models—Linear Regression (LR), M5P, Multi-layer Perceptron (MLP), Sequential Minimal Optimization – Support Vector Machine (SMO-SVM), Random SubSpace (RSS), Random Forest (RF), and Random Tree (RT)—in predicting the green and blue water footprints of rice in Punjab, India. Best subset regression and correlation matrix indicate that humidity, wind speed, sunshine hours, solar radiation, and total rainfall are optimal inputs for green water footprint prediction, while maximum temperature, humidity, wind speed, sunshine hours, and solar radiation are best for blue water footprint prediction. The RT model outperformed others in that, for green water footprint prediction, it achieved a correlation coefficient (CC) of 0.9991, mean absolute error (MAE) of 0.1314, root mean square error (RMSE) of 0.4553, relative absolute error (RAE) of 0.0477, and root relative squared error (RRSE) of 0.1283 during the training stage. However, during the testing stage, the RF model performed better (CC = 0.79, MAE = 154.2732, RMSE = 192.3973, RAE = 55.5602, and RRSE = 58.6433). For blue water footprint prediction, the RT model remained the best performer in both stages (training: CC = 0.9991; testing: CC = 0.9981, MAE = 0.7920, RMSE = 0.8583, RAE = 0.4440, and RRSE = 0.8290). These results suggest that machine learning can effectively support water management strategies by providing quick and reliable estimates of water footprints, which is crucial for sustainable rice production. By utilizing these models, policymakers can make informed decisions to optimize water usage and ensure sustainable agricultural practices.