Hybrid CNN & Random Forest Model for Effective Bitter Orange Leaf Disease Diagnosis

This study focuses on the right category of illnesses of the leaves in the bitter orange leafplants using modern machine learning approaches. This research employs the Convolutional Neural Network (CNN) or Random Forest models on a rich dataset of photos of bitter orange leaves damaged by the various diseases. The model's performance is carefully evaluated using precision, recall, or F1-Score measures, allowing for a thorough study of its ability to distinguish between various leaf diseases. The results show that the model is accurate in the positive predictions, with precision values that range from 92.85% to 95.48%. High recall values ranging from 92.61% to 94.82% demonstrate its capacity to record actual incidences of each disease of the leaf. The F1-Score, a beautiful blend of the recall and accuracy, ranges from 93.11% to 94.67%, demonstrating balanced performance across distinct illness classes. The prevalence of every leaf disease within the dataset is revealed by the support values, providing significant information for handling the diseases in actual agricultural situations. The model's general success across multiple illness classes is reinforced by the macro-average, weighted on a typical basis, or micro-average metrics, which are constantly around 94.06%. The obtained accuracy of the 98% demonstrates the model's ability to accurately classify the majority of occurrences. These findings make an important contribution to the field of agricultural and the precision by providing a dependable method for early identification and intervention in citrus farming. The complete evaluation approach guarantees the model's resilience in the practical scenarios, promoting long-term citrus leaf disease control practices.