Predictive Modeling for Crop Health and Recommendation: Leveraging Machine Learning Algorithms to Optimize Agricultural Sustainability

Efficient and sustainable harvest management is crucial to addressing the global challenges of nutritional safety. This study examines the application of predictive modeling techniques to optimize plant health management and to predict the types of harvests that can grow based on a variety of ecological and soil-related factors. Various algorithms for machine learning were trained and evaluated using data records with plant-related features including soil composition, precipitation, pH, and macronutrient content. Models include Random Forests, Gradient Boosting, AdaBoost, Support Vector Machines (SVMs), Logistics Regression, K-Nearest Neighbor (KNN), decision trees and Artificial Neural Networks (ANNs). Random Forest showed the highest accuracy (99.32%) and F1 score (0.993), followed by a Decision Tree (98.64%) and Gradient Boosting (98.18%). In contrast, AdaBoost was significantly below average (accuracy: 14.09%, F1 score: 0.085). These results show that ensemble methods, particularly Random Forest, are highly effective in predicting harvests for different environments and identifying the optimal harvest type. This study highlights the capability of machine learning in the further development of precision agriculture by providing a strong data control solution for harvest management. Future work will consider real-time implementation and integration into IoT devices to improve campaign scalability and applicability.