Scholarlink

A key element of life for humans is agriculture. Not only is it essential, but it contributes to the development of a nation. For a substantial percentage of Indians, agriculture is a significant source of income. Agriculture and Rural Development provide a substantial amount of Indians’ principal revenue source income. In recent years, ecological damage has become an important concern for people all around the world. It has been found that ecological change has a consequence on farming quality and quantity. Crop yield forecasts would help producers make informed marketing and storage decisions well before harvest. Many landowners anticipate using old-fashioned agricultural methods, which produce little profit in today's competitive marketplace. Agricultural forecasting of yields is critical for quick decision-making both on the national and regional levels. An accurate crop production forecasting model could help in deciding which to plant and how much yield can be produced. The implementation of Machine Learning (ML) is an important aspect to consider when trying to find a useful and efficient solution to the crop yield problem. Crop yield can be predicted by analyzing different datasets given and various ML algorithms like Linear Regression, Random Forest (RF) Regression, Support Vector Regression (SVR), Decision Tree (DT) Regression and Extreme Gradient Boosting (XGBoost) Regression. The feature selection methods include ANOVA, the correlation-based threshold value method, the Mutual information method, and the Pearson correlation methods used for three different datasets. The accuracy of the DT algorithm in the agricultural yield prediction dataset has a highest accuracy of 97.25% after different feature selection techniques are used. With a 99.2% accuracy rate in the dataset used to forecast wild blueberry yield, the Linear Regression approach is the most accurate. The highest precision of the RF algorithm is 93.53%, and that of the XGBoost technique is 94.28% in the Rice Production Prediction Dataset using ANOVA and correlation coefficient threshold feature selection methods, respectively.

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

Feature Selection Strategies for Accurate Crop Yield Prediction in Heterogeneous Datasets

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • R. Ramani,
  • M. Uma Maheswari,
  • S. Kayathri

摘要

A key element of life for humans is agriculture. Not only is it essential, but it contributes to the development of a nation. For a substantial percentage of Indians, agriculture is a significant source of income. Agriculture and Rural Development provide a substantial amount of Indians’ principal revenue source income. In recent years, ecological damage has become an important concern for people all around the world. It has been found that ecological change has a consequence on farming quality and quantity. Crop yield forecasts would help producers make informed marketing and storage decisions well before harvest. Many landowners anticipate using old-fashioned agricultural methods, which produce little profit in today's competitive marketplace. Agricultural forecasting of yields is critical for quick decision-making both on the national and regional levels. An accurate crop production forecasting model could help in deciding which to plant and how much yield can be produced. The implementation of Machine Learning (ML) is an important aspect to consider when trying to find a useful and efficient solution to the crop yield problem. Crop yield can be predicted by analyzing different datasets given and various ML algorithms like Linear Regression, Random Forest (RF) Regression, Support Vector Regression (SVR), Decision Tree (DT) Regression and Extreme Gradient Boosting (XGBoost) Regression. The feature selection methods include ANOVA, the correlation-based threshold value method, the Mutual information method, and the Pearson correlation methods used for three different datasets. The accuracy of the DT algorithm in the agricultural yield prediction dataset has a highest accuracy of 97.25% after different feature selection techniques are used. With a 99.2% accuracy rate in the dataset used to forecast wild blueberry yield, the Linear Regression approach is the most accurate. The highest precision of the RF algorithm is 93.53%, and that of the XGBoost technique is 94.28% in the Rice Production Prediction Dataset using ANOVA and correlation coefficient threshold feature selection methods, respectively.