An Explainable Dynamic Ensemble Selection Model for Interpretable Almond Type Classification

The rich source of vitamins and nutrients makes almonds crucial for maintaining the health of the personnel, in addition to their high commercial importance. To preserve the quality of almonds, it is essential to grade and classify them correctly. Recent techniques have used standard machine learning algorithms for classification; however, these often fail to provide robust results. Furthermore, the studies do not explain the outputs generated by these models, which is crucial in understanding the behavior of the models. Therefore, in this research, the authors have implemented techniques of Dynamic Ensemble Selection (DES), consisting of K Nearest Oracles Eliminate (KNORA-E) and K-Nearest Oracles Union (KNORA-U), for the classification of almonds. It was observed that these techniques have outperformed static ensemble classifiers, achieving the highest accuracy of 85% to 87%, respectively. Further, methods of explainable AI, namely Local Interpretable Model Agnostic Explanation (LIME) and Shapley Additive Explanations (SHAP), have been implemented to understand the contribution of features in the model’s decision-making, thereby assisting in interpreting the correctness of the models