Hybrid CNN-Random Forest Framework for Elevated Accuracy in Powdery Mildew Disease Classification

This research paper's abstract provides a thorough evaluation of the predictive capability of a model developed to classify the mango dusty leaf disease. The study uses an evaluation methodology that includes criteria like Precision, Memory, F1-Score, Support, Supported Proportion, and Accuracy and focuses on different disease Seriousness Levels. The evaluation shows the model's capacity to distinguish and categorize various disease Seriousness levels, ranging from Level 1 down to Level 10. While recall, which measures the capacity to properly recognize real positives, ranges between 92.05% and 91.43%, precision, which measures the accuracy in making optimistic predictions, goes from 93.80% to 92.29%. The F1-Score ranges from 97.62% to 99.65% and balances Precision and Recall. Additionally, the study quantifies that Support, showing the frequency of each class and providing information on the prevalence of various disease Seriousness Levels. The Support Proportion illustrates how examples differ among Seriousness Levels, which helps to explain class disparities. Precision, Recall, and F1-Score had macro-average values of 97.56%, 93.84%, and 96.35%, respectively, calculated over all classes. The weighted average measures significantly noticeably higher, displaying 95.40% for accuracy 96.00% for Remember, and 94.90% for F1-Score, while also taking class imbalances and consistency into account. The micro-average, which averages the performance over all instances, displays accuracy, recall, and F1-Score values of 96.77%, 96.49%, and 96.63% respectively. The results show how well the model categorizes the severity of mango powdery leaf disease.