AI in Agriculture: Hybrid Convolutional Neural Networks and Random Forest for Sunflower Disease Identification

To control sunflower diseases, their identification and categorization appeal to increase agricultural yield and minimize producers’ losses. This paper presents a new model that comprises CNN for feature extraction and RF for classification with the goal of early diagnosis of sunflower diseases. The model was evaluated on a dataset of 5,000 labeled sunflower leaf images across five classes: Twig and fruit blights, mildews, rusts, leaf spots, and mosaic virus. After 30 epochs of the training phase, the proposed hybrid model got 99.00% training accuracy and a validation accuracy of 97.00%, showing that the developed pattern can effectively generalize the data without any chance of overfitting. Accuracy is comparable across all datasets, with an average precision of 0.9650, recall of 0.9630, and F1-Score of 0.9640. The confusion matrix provides nearly no misclassifications in addition to a high average Area Under the Curve (AUC) of 0.98, which shows the model’s discriminant feature. Scatter plots and 3D feature clustering provided a better view. They showed that the disease classes are separable, providing merit for CNN in feature extraction—further, the incorporation of RF-enabled accurate classification regardless of class imbalance. As a result, this study highlights the usability of the proposed hybrid CNN-RF model as a valuable instrument for practical use in precision agriculture, which would allow farmers to apply accurate and suitable disease management measures in time. These conclusions may be helpful for future research that substantiates the versatility, flexibility, and relative use of the developed technique for other crops and other agricultural problems.