Unveiling Predictive Factors in Apple Quality: Leveraging LIME, SHAP, and the Synergy of Machine Learning Models and Artificial Neural Networks

In the realm of fruit quality assessment, the demand for accurate identification methods has grown significantly. Traditional measures, such as physical and chemical analyses, prove impractical due to their time-consuming nature and potential sample damage. To address this imperative need for precise and timely apple quality detection, this research explores the application of machine learning algorithms. Recognizing the drawbacks of traditional methods reliant on physical and chemical measures for predicting fruit maturity, non-destructive techniques, such as hyperspectral imaging and machine learning algorithms, are meticulously investigated. The study goes on to evaluate and compare the performance of various machine learning approaches, including Random Forest Classifier (RF), K-Nearest Neighbors Classifier (KNN), Support Vector Classifier (SVC), Extreme Gradient Boosting Classifier (XGB), Light Gradient Boosting Machine Classifier (LGBM), Artificial Neural Network (ANN), and Voting Classifier. Among these, the Artificial Neural Network emerges as the top performer, achieving an impressive accuracy rate of 92.87%. Following the thorough evaluation of the models, Explainable Artificial Intelligence (XAI) techniques, specifically LIME and SHAP, are employed. These techniques playa crucial role in interpreting and visualizing the intricate decision- making process of the models. The insights gained underscore the models' robustness while providing valuable information on the decision-making processes, contributing significantly to advancements in apple quality detection methodologies.