Scholarlink

<p>Efficient nutrient management remains a major challenge in precision agriculture due to spatial variability in soil properties and the need to balance crop productivity, economic cost, and environmental sustainability. This study proposes a multi-objective decision support system (MODSS) for zone-specific fertilizer recommendation by integrating machine learning-based soil fertility prediction, spatial clustering, and evolutionary optimization. The framework utilizes heterogeneous inputs, including geospatial attributes, vegetation indices (NDVI, EVI, SAVI), and soil moisture, to model soil variability and predict fertility levels. Among the evaluated models, the Random Forest algorithm achieved the best performance with an accuracy of 92.4%, precision of 0.92, recall of 0.90, and F1-score of 0.91, along with low prediction errors (RMSE: 6.8&#xa0;kg/ha and MAE: 5.2&#xa0;kg/ha). Spatial variability was effectively captured using KMeans clustering, where the optimal number of zones was identified as k = 5 based on a Silhouette Score of 0.63, Davies-Bouldin Index of 0.98, and Calinski-Harabasz Score of 610.8. This enabled accurate delineation of management zones for targeted nutrient application. The predicted soil conditions were integrated into an NSGA-II-based multi-objective optimization model to balance yield, cost, and environmental impact. The optimal solution achieved a yield of 4.5 t/ha with a cost of ₹11,000/ha and controlled environmental impact. Compared to traditional methods, the proposed system improved yield by 18.4%, reduced fertilizer cost by 13.6%, and decreased nutrient usage by up to 25%.</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

A multi-objective decision support system for zone-specific fertilizer recommendation using soil fertility prediction models

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月28日

查看全文
  • R. Lakshmi,
  • J. Vijayashree

摘要

Efficient nutrient management remains a major challenge in precision agriculture due to spatial variability in soil properties and the need to balance crop productivity, economic cost, and environmental sustainability. This study proposes a multi-objective decision support system (MODSS) for zone-specific fertilizer recommendation by integrating machine learning-based soil fertility prediction, spatial clustering, and evolutionary optimization. The framework utilizes heterogeneous inputs, including geospatial attributes, vegetation indices (NDVI, EVI, SAVI), and soil moisture, to model soil variability and predict fertility levels. Among the evaluated models, the Random Forest algorithm achieved the best performance with an accuracy of 92.4%, precision of 0.92, recall of 0.90, and F1-score of 0.91, along with low prediction errors (RMSE: 6.8 kg/ha and MAE: 5.2 kg/ha). Spatial variability was effectively captured using KMeans clustering, where the optimal number of zones was identified as k = 5 based on a Silhouette Score of 0.63, Davies-Bouldin Index of 0.98, and Calinski-Harabasz Score of 610.8. This enabled accurate delineation of management zones for targeted nutrient application. The predicted soil conditions were integrated into an NSGA-II-based multi-objective optimization model to balance yield, cost, and environmental impact. The optimal solution achieved a yield of 4.5 t/ha with a cost of ₹11,000/ha and controlled environmental impact. Compared to traditional methods, the proposed system improved yield by 18.4%, reduced fertilizer cost by 13.6%, and decreased nutrient usage by up to 25%.