The impact of duty ratio-controlled ion energy on the surface roughness of silicon nitride (SiN) films is examined as a function of the bias power and the duty ratio. Experimental ranges are 30-90 W and 20-80% for the bias power and the duty ratio, respectively. SiN films were deposited at room temperature using a SiH4-NH3 pulsed plasma. Atomic force microscopy was used to measure surface roughness. Investigation is detailed in view of a mean-surface roughness, a non-uniformity of surface height distribution, and using a prediction model. The prediction model was constructed using a neural network and a genetic algorithm. A decrease in the duty ratio results in an increase in the surface roughness, but a decrease in the non-uniformity. Correlation study revealed that the surface roughness was strongly related to the ion energy and the ion energy flux. The neural network model predicts the high ion energy as the most influential diagnostic parameter.