Wildthing

I built a learning model (for classification) based on a Random Forest classifier and i am asked to assess the statistical significance of its performances. Up to now, i trained and tested it on two different datasets A and B, respectively. What kind of test can i use?