Explainable AI with SHAP for Data-Driven Growth Prediction in Smart Poultry Farming

Smart poultry farming increasingly leverages Artificial Intelligence (AI) and the Internet of Things (IoT) to improve monitoring and decision-making. This paper presents an explainable predictive framework for broiler growth using IoT-based farm data that combine environmental measurements with management variables (feed, water, age, mortality). We benchmark Random Forest (RF), Support Vector Regression (SVR), and Extreme Gradient Boosting (XGBoost) under a time-aware evaluation protocol, and assess performance using RMSE, MAE, and R2. RF delivers the best accuracy (RMSE = 21.56, MAE = 16.84, R2 = 0.997), outperforming SVR and XGBoost. To ensure transparency, SHapley Additive exPlanations (SHAP) are used to analyze model behavior: feed, water, and age emerge as primary positive contributors, whereas elevated ammonia (NH3), methane (CH4), and hydrogen sulfide (H2S) are associated with negative effects. These results indicate that the approach provides both high predictive performance and actionable insights for management, supporting trustworthy decision support in smart poultry systems. Future work will extend validation to multifarm, multi-season datasets and explore real-time edge/cloud deployment for on-farm use.