A lightweight selective stacking framework for IoT crop recommendation

Abstract

The integration of the internet of things (IoT) into precision farming has revolutionized agricultural practices by enhancing resource management and crop productivity through real-time data monitoring and automation. Precision farming leverages IoT technologies to monitor critical environmental factors such as soil moisture, temperature, humidity, and nutrient levels, enabling farmers to make data-driven decisions for optimizing crop growth. Despite these advancements, improving the accuracy and efficiency of IoT-based crop recommendation systems remains a key challenge, particularly in resource-constrained environments. This study aims to enhance the predictive performance of IoT-based crop recommendation systems by developing a novel stacked deep ensemble learning model. The proposed model, termed lightweight selective stacking with deep ensemble learning (LSSDEL), focuses on reducing computational complexity while maintaining high predictive accuracy. Key methods employed include selective model stacking, L1 regularization for model pruning, gradient-free model aggregation, and the implementation of an early stopping mechanism. The system is validated using real-world IoT agricultural datasets, emphasizing its scalability and practical applicability. Findings from the study demonstrate that the LSSDEL model outperforms traditional models, achieving a prediction accuracy of 97.80% and significant improvements in precision, recall, and F1-score. Furthermore, the proposed model reduces execution time by 16.7% compared to existing approaches, confirming its computational efficiency.