below are two models of random forest (RF). In the first one I used cross validation (CV) to validate my model. In the other one RF is used without any validation method directly on the training data. In both cases the same training data set, as well as the same real data set to apply the model is used. However the performance of both differs (slightly).
My question is: Shouldnt performance be the same?
If I use some other algorithm e.g. DT,NB, performance of both models (with and without CV) are the same.