Scholarlink

<p>Modern farming benefits from precision agriculture because it enables data-based choices which make operations more sustainable and efficient. The produced recommender system utilizes ensemble methods to provide optimized guidance regarding crop selection as well as fertilizer recommendations. The system combines Random Forest, Gradient Boosting and Support Vector Machines with K-Nearest Neighbors and SGD Classifier and XGBoost algorithms to train using SMOTE which balances class distribution in the datasets. From achieved results which show 100% accuracy, precision, recall and F1-score in both prediction tasks indicate the model’s strong capacity to apply in real agricultural scenarios. Through the usage of LIME (Local Interpretable Model-Agnostic Explanations) decision makers gain clarity about how the AI system functions by revealing that crop selection depends critically upon moisture and phosphorus levels and how soil type and potassium levels are strong determinants for fertilizer recommendations. Despite the promising results, the use of synthetic oversampling introduces the risk of overfitting. Accordingly, the study acknowledges the need for further validation using robust feature selection strategies, hyperparameter tuning, and cross-evaluation techniques. Additional studies will examine the field practicality of deploying solutions between different farming locations to evaluate long-term operational sustainability. This research demonstrates how interpretable AI provides farmers with exact sustainable agricultural insights which they can directly implement.</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

Unveiling Sustainable Agriculture: An Ensemble Approach for Precision Crop and Fertilizer Recommendation with Local Interpretable Model-Agnostic Explanations

  • 期刊论文

  • 发表日期：

    2025年10月23日

查看全文
  • Chetan R,
  • D V Ashoka,
  • Ajay Prakash B V,
  • Basamma Umesh Patil

摘要

Modern farming benefits from precision agriculture because it enables data-based choices which make operations more sustainable and efficient. The produced recommender system utilizes ensemble methods to provide optimized guidance regarding crop selection as well as fertilizer recommendations. The system combines Random Forest, Gradient Boosting and Support Vector Machines with K-Nearest Neighbors and SGD Classifier and XGBoost algorithms to train using SMOTE which balances class distribution in the datasets. From achieved results which show 100% accuracy, precision, recall and F1-score in both prediction tasks indicate the model’s strong capacity to apply in real agricultural scenarios. Through the usage of LIME (Local Interpretable Model-Agnostic Explanations) decision makers gain clarity about how the AI system functions by revealing that crop selection depends critically upon moisture and phosphorus levels and how soil type and potassium levels are strong determinants for fertilizer recommendations. Despite the promising results, the use of synthetic oversampling introduces the risk of overfitting. Accordingly, the study acknowledges the need for further validation using robust feature selection strategies, hyperparameter tuning, and cross-evaluation techniques. Additional studies will examine the field practicality of deploying solutions between different farming locations to evaluate long-term operational sustainability. This research demonstrates how interpretable AI provides farmers with exact sustainable agricultural insights which they can directly implement.