Random forest based approach for crop recommendation

Abstract

The selection of crop species that are well-suited to the soil conditions, climate, and water availability is essential for maximizing resource efficiency, boosting yields, and supporting food security. Traditional methods for crop recommendation involve empirical methods and statistical analysis of large amounts of data, which limits the capacity to respond effectively to different challenges. The use of machine learning algorithms for crop recommendation has become a key tool to support farmers’ decision-making. In this work, we propose an effective crop recommendation model based on the Random Forest algorithm, which is optimized for this agricultural context. The model is trained with inputs corresponding to soil composition and climate data, while the output corresponds to the best crop for a given input condition. The proposed methodology involves preprocessing the data and finding the best hyperparameters for the random forest-based classification. Additionally, a thorough evaluation of the resulting model is implemented. The results of this methodology show an improvement in the accuracy, precision, recall, and F1-score metrics for the recommendation of 22 different crop classes in comparison with other classical machine learning and state-of-the-art approaches. This work demonstrates that our approach can provide a practical and accessible solution in systems within agricultural environments with limited infrastructure.