During the spent fuel reprocessing process, nitrogen oxides (NOx) gases are generated. The treatment and emission control of NOx rely on measurements from instrumentation. In situations where operating conditions fluctuate, the response capability of the treatment system exhibits a lag, resulting in a rapid short-term increase in NOx concentration during final emissions. To predict the trend of NOx concentration changes in the reprocessing process and enhance the response capability of the NOx treatment system, a NOx concentration prediction model was developed using the Random Forest algorithm, based on data collected from actual operations. Feature engineering was employed to select variables, further improving the model's R2 to 0.92. The results indicate that the Random Forest model demonstrates excellent predictive performance for NOx concentration data and outperforms traditional machine learning models.