Hybrid Ensemble Model for Accurate Orange Quality Prediction Using CNN, Random Forest, and Gradient with Multispectral Data Integration

This work presents a Hybrid Ensemble Learning Model for predicting orange quality, integrating image-based and structured data to achieve a comprehensive assessment of fruit ripeness and quality. The proposed model combines Convolutional Neural Networks (CNN) for extracting visual features from high-resolution orange images with Random Forest and Gradient Boosting algorithms for processing structured data, including sweetness (Brix), pH, size, and environmental conditions. This hybrid architecture capitalizes on CNN's visual feature extraction capabilities alongside the robustness of Random Forest and Gradient Boosting in handling structured data, enhancing classification precision. A meta-classifier is employed to combine the outputs from these individual classifiers, achieving a seamless fusion of decisions from the diverse data sources. The model was tested on a dataset with multiple quality levels, aiming to optimize harvest timing and improve agricultural decision-making. With an achieved accuracy of 91%, the model demonstrates significant potential for application in automated, data-driven fruit quality prediction, reducing dependency on manual inspection and increasing consistency. This Hybrid Ensemble Learning approach represents an advancement in precision agriculture, highlighting the benefits of multi-source data fusion in predicting fruit ripeness and quality in citrus production.