Ensemble RF-KNN Model for Accurate Prediction of Drought Levels

Abstract

The increasing frequency and severity of droughts represent a critical threat to agricultural systems worldwide, disrupting food production, and supply chains. Accurate and timely prediction of drought conditions is essential for ensuring agricultural sustainability and enabling proactive mitigation strategies. This study proposes a novel ensemble model that combines Random Forest (RF) and K-Nearest Neighbors (KNN) using soft voting to predict drought conditions based on meteorological data. The dataset consists of drought classifications for six levels, ranging from no drought to five drought severity levels using meteorological indicators from various U.S. counties. The performance of the proposed model was evaluated against several state-of-the-art machine learning models, including Logistic Regression, Decision Tree, and Artificial Neural Networks, using various evaluation metrics including accuracy, precision, recall, and F1-score. The results demonstrate the effectiveness of the proposed ensemble approach, achieving superior accuracy and reliability in predicting drought severity. This research highlights the transformative potential of machine learning in supporting agricultural systems and addressing climate change challenges through data-driven drought monitoring and mitigation strategies.