Comparison and Evaluation of Machine Learning Models for Predicting CH4 Emissions from Pig Housing

CH is a major air pollutant emitted from pig housing and is influenced by various environmental and management factors.This study monitored CH 4 concentrations using sensor-based measurements and analyzed their relationships through machine learning models.Outliers were detected using the Hampel filter, and missing values were corrected with hybrid imputation techniques to ensure data quality.Among Random Forest, LightGBM, and HistGB models, the HistGB model showed the highest predictive accuracy (R 2 = 0.83) and lowest RMSE.The results indicate that feed intake, pH, and temperature significantly affect CH emissions, while ventilation and humidity primarily influence NH 3 variation.This study provides a data-driven approach for identifying key emission factors and improving gas management in intensive pig production systems.