Ammonia synthesis is vital for fertilizer production, but the traditional Haber-Bosch process is energyintensive and environmentally burdensome due to its high-temperature and high-pressure operations. Plasma-catalytic ammonia synthesis offers a sustainable alternative, generating large datasets under various experimental conditions. To optimize energy efficiency, we established a database with 305 data points and 7 experimental parameters, each linked to its corresponding energy efficiency. We employed an Extreme Gradient Boosting (XGBoost) regression tree model, achieving an average R² value of 0.9434 for predictions. Bayesian Optimization (BO), using Gaussian Process Regression as a surrogate model, systematically explored the experimental parameter space. It utilized XGBoost predictions to identify parameter combinations that maximized energy efficiency. After 50 iterations, the optimal parameters were identified: 6.4 g catalyst mass, 50 mm grounding electrode length, nickel metal catalyst, Al₂O₃ catalyst support, 5 W power, 160 ml·min⁻¹ flow rate, and a 1:2 feed ratio. Under these conditions, the energy efficiency of plasma-catalytic ammonia synthesis improved to 1.49 g·kW·h⁻¹, a 22.1% increase from the highest value of 1.22 g·kW·h⁻¹ in the dataset.