An IoT-Driven hybrid AI model for health monitoring of cows

Livestock diseases continue to pose significant challenges for smallholder dairy farmers, particularly in regions with limited access to veterinary services and real-time health monitoring infrastructure. To address this, the present study proposes a cost-effective IoT-enabled framework for early disease detection in dairy cattle, targeting smallholder farmers in resource-limited regions. A custom smart collar was developed to monitor body temperature, pulse rate, and activity levels in 150 cows across seven districts of Punjab. To analyse the collected data, a novel hybrid model, SM-GBoost-LSTM, was implemented, combining Gradient Boosting (GBoost) for structured features, Long Short-Term Memory (LSTM) networks for temporal patterns, and SMOTE for handling class imbalance. The model achieved high performance, attaining 93.56% accuracy, 91.42% precision, 77.77% recall, and 84.02% F1-score. The system also integrates real-time cloud analytics and a bilingual mobile application (English/Punjabi) to deliver timely health alerts. Overall, the proposed framework provides a scalable, field-validated, and affordable AI-driven solution for improving livestock healthcare in smallholder dairy farming. Proposes a cost-effective IoT-enabled smart collar design for continuous monitoring of key physiological parameters in dairy cattle. Introduces SM-BoostLSTM, a hybrid model integrating SMOTE (for class imbalance), Gradient Boosting (for structured feature learning), and LSTM (for temporal pattern recognition) in livestock health data. Demonstrates the efficacy of the framework through field-validated analysis and a bilingual (English–Punjabi) mobile app envisioned for delivering timely health alerts to farmers.