Combining CNN and SVM for Robust Cattle Disease Classification in Veterinary Applications

Noteworthy problems in the cattle livestock sector stem from infectious diseases that impact the health of animals and the production output. Precise and prompt detection is essential for both treatment and prevention to be successful. This study presents a CNN-SVM model to enhance the exactness of categorizing cattle diseases. The dataset included 2,000 photographs put into five illness classifications (Healthy, Foot-and-Mouth illness, Mastitis, Bovine Respiratory Disease, and Lameness), with 80 % for training and 20% for evaluation. CNN turned to ImageNet, which turned out to be a feature extractor, and the SVM with a radial basis function (RBF) kernel categorized the identified features. Experimental results reveal that the hybrid model outperforms the independent CNN and SVM models. Ultimately, the success rate of the CNN-SVM model is 92.7 %, more successful than the CNN (88.4%) and SVM models (84.5%). Also, the hybrid model attained an average accuracy of 91.5 %, a recall of 90.8 %, and an F1 score of 91.1%, showing a balanced performance concerning all criteria. The figure of 0.93 for the Area Under the Curve (AUC) of the Receiver Operating Characteristic (ROC) curve for all illness categories, on average, suggests the model has outstanding discriminative capabilities. The evidence reveals that the CNN-SVM hybrid model is a valid and valuable classification method for recognizing diseases in cattle, which stands out as an essential development from other techniques. Additional research is required to better the dataset and examine alternative model frameworks to elevate performance and relevance under actual circumstances.