PREDICTIVE DIAGNOSIS OF RUMINANT DISEASES USING ARTIFICIAL INTELLIGENCE

Abstract

This study proposes an effective and robust framework using AI, based on the multi-modal datasets of real-life dairy and mixed-farming conditions, that help in diagnosing ruminant diseases predictively.  Over 8,000 ruminants were used to collect data in terms of sensor-gathered physiological data, laboratory test parameters, case histories, and ambient properties.  It successfully detected the illnesses such as mastitis, respiratory infections, and ruminal acidosis because it used a hybrid type of modelling approach that combined XGBoost and recurrent neural networks.  This was an optimized model with the score of 0.89 F1 and 91.4% diagnostic accuracy.  The rates of ruminations, temperature alteration, and pH were considered to be meaningful predictors according to the explanations of features through SHAP values.  A mobile and web-based model implementation made field-level, real-time diagnostic alerts possible and they were later enhanced through a veterinary feedback loop.  This literature demonstrates how AI can be applied in optimising the outcome of animal care in precision livestock farming systems, reducing losses in terms of financial expenses and an increase in detecting illness at an early stage.