Enhancing Crop Yield Prediction through Random Forest Classifier: A Comprehensive Approach

Abstract

Accurate crop yield prediction is pivotal in precision agriculture, enabling optimized resource allocation and enhanced decision-making for farmers. Despite significant advancements in machine learning (ML) models, existing approaches often lack interpretability and insufficient integration of diverse environmental data. This study fills in these gaps by creating a strong crop recommendation system using a random forest classifier (RFC). It achieves this by utilizing the RFC’s ensemble nature to improve prediction accuracy and understanding of which features are most important. The proposed model works with a dataset that includes environmental factors and soil nutrients. It does this by using the interquartile range (IQR) method to find outliers and logarithmic transformations to fix skewed data. The RFC’s capability to manage high-dimensional data and prevent overfitting is validated through rigorous evaluation, achieving an impressive accuracy of 99.13% on the testset. This performance surpasses other non-linear models, such as Gaussian Naive Bayes (GNB) and XGBoost, which attained accuracies of 99.2% and 98.5%, respectively. The model’s high accuracy and interpretability offer a practical real-time crop recommendation tool, facilitating sustainable agricultural practices. This research contributes to the field of crop yield prediction and sets a precedent for integrating ML with practical agricultural applications.