Loading content...
Plant diseases significantly reduce agricultural productivity and crop quality worldwide. This study presents an explainable hybrid ensemble learning framework for plant disease prediction and crop improvement. The proposed framework integrates Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Random Forest (RF), and Convolutional Neural Network (CNN) models to enhance predictive robustness and classification performance. In addition, explainability techniques, including SHAP and LIME, are employed to interpret model predictions and improve transparency. Experimental results demonstrate superior performance compared with individual models, achieving an accuracy of 97.9%, precision of 97.4%, recall of 97.1%, F1-score of 97.2%, and AUC of 99.1%. The proposed approach effectively improves disease detection across diverse crop types while reducing misclassification rates. By combining high predictive accuracy with model interpretability, the framework supports early disease diagnosis, informed decision-making, and sustainable agricultural practices, thereby contributing to precision agriculture and enhanced crop productivity.
Plant diseases, which can drastically lower crop output and quality, continue to be a major obstacle to agriculture, which is nevertheless essential to both global food security and economic stability. Conventional disease detection techniques mostly rely on expert knowledge and manual field inspections[1], which are frequently labor- intensive, time-consuming, and prone to human mistake. As a result, there is a growing need for sophisticated computer methods that can precisely forecast plant diseases in their early stages, allowing for prompt intervention and crop enhancement. Plant disease diagnosis[2] using image-based data has shown great potential thanks to recent developments in machine learning (ML) and deep learning (DL) approaches. From visible indications like leaf discoloration, lesions, and morphological alterations, these algorithms are able to extract intricate patterns. However, the practical use of most current models in actual agricultural settings is constrained by their limitations in terms of interpretability, generalization across crop kinds, and robustness under various environmental conditions. Hybrid ensemble learning techniques have become a viable way to overcome these constraints[3]. Ensemble approaches improve model stability, decrease overfitting, and increase overall accuracy by merging several predictive models. To capture both geographical and temporal aspects of plant disease progression, hybrid ensembles combine the advantages of many classifiers, including decision trees, support vector machines, convolutional neural networks, and recurrent neural networks. Furthermore, explainable artificial intelligence (XAI)[4] methods have become popular because they make model predictions transparent, allowing agronomists and farmers to comprehend the underlying causes of disease diagnosis. In agriculture, where the adoption of automated systems depends on the results' interpretability and trustworthiness, explainability is very important.