Buffalo Disease Diagnosis Using Machine Learning: A Symptom-Based Text Classification Approach

Abstract:Background: Buffaloes play a crucial role in the agricultural economy, especially in regions dependent on dairy and draught animals. However, research specifically targeting disease detection in buffaloes remains limited despite their susceptibility to several infectious diseases. Early and accurate diagnosis is vital for managing disease outbreaks and ensuring herd health. This study uses machine learning (ML) and deep learning (DL) models to emphasize buffalo-specific disease classification. Five commonly occurring diseases, anthrax, blackleg, foot and mouth disease, lumpy skin disease, and pneumonia, were investigated using symptom-based textual descriptions, focusing on enhancing diagnostic accuracy for buffaloes. Methods: Textual symptom data were collected and pre-processed using Term Frequency-Inverse Document Frequency (TF-IDF) to convert unstructured text into numerical feature representations. The study explored three different classification algorithms: Multilayer Perceptron (MLP), Convolutional Neural Network (CNN), and XGBoost. Each model was trained and evaluated on species-specific subsets, with particular attention given to buffalo disease data. Performance was measured using classification accuracy and disease-wise detection effectiveness to assess the suitability of each model for buffalo diagnostics. Results: MLP consistently outperformed the other models in classifying diseases in buffaloes, particularly for anthrax and blackleg, which exhibit distinct symptoms. CNN demonstrated robust handling of complex symptom patterns, while XGBoost provided stable and generalized results. However, the classification accuracy declined for diseases with overlapping clinical features, such as pneumonia and lumpy skin disease. These patterns highlight the challenges in differentiating symptomatically similar diseases and indicate the need for enhanced symptom representation in future research. Conclusion: Based on textual symptom data, the study demonstrates the feasibility and effectiveness of using ML and DL models for automated disease classification in buffaloes. MLP, in particular, shows promise for integrating into intelligent decision-support tools to improve diagnostic accuracy and response time in Buffalo Healthcare. The findings contribute to species-specific veterinary informatics and support the development of targeted surveillance systems for managing buffalo health more effectively.