Prediction of Plant Disease Severity Using Advanced Gradient Boosting Techniques

Abstract

Plant diseases significantly impact agricultural productivity, causing economic losses and food insecurity. Predicting the severity of plant diseases is crucial for timely interventions and sustainable farming practices. However, existing methods often lack precision and scalability. This study aims to develop a highly accurate model to predict plant disease severity using advanced gradient boosting techniques. We used the publicly available New Plant Diseases Dataset from Kaggle, containing images of healthy and diseased plants. Data preprocessing included image augmentation and feature extraction using transfer learning. Gradient boosting algorithms such as XGBoost and LightGBM were employed to build predictive models. The proposed model achieved an accuracy of 96.45% on the test set, outperforming existing approaches. The results demonstrate the model's robustness in predicting disease severity across multiple plant species. This study highlights the potential of gradient boosting techniques for precise plant disease severity prediction, offering a scalable solution for agricultural applications.