Ensemble learning based plant disease prediction and analysis: A comparative study

Crop illnesses considerably affect agricultural productivity and food security, making prompt and precise identification essential to minimize losses and support sustainable agriculture. This research investigates the performance of deep learning models versus ensemble approaches for detecting plant diseases within the PlantVillage dataset. A Convolutional Neural Network (CNN) which is on MobileNetV2 was utilized in feature extraction, and it was compared to standalone classifiers - XGBoost, Support Vector Machine (SVM), and Random Forest - as well as their ensemble. The research assesses the predictive performance of these models, emphasizing how the ensemble can merge strengths and minimize misclassification. Experimental findings indicate that the ensemble model reaches an accuracy of 94.1%, surpassing individual models (CNN: 92.5%, Random Forest: 88.3%, SVM: 85.6%, XGBoost: 89.4%). This comparative study delivers insights into the trade-offs of models, presenting a scalable approach for automatic detection of plant diseases in precision farming.