"x-validation vs. performance operator"

fmon

Hi
I am designing a prediction model based on decision tree.
Once I used an x-validation operator consisting of decision tree in training part and apply model and performance in testing part.Another time I used a sequence of three operators of decision tree,apply model and performance.But I got different results.I'd like to know the reason.
Thanks

fras

To get a roughly performance error estimation only the first method is suitable (X-Validation).
In your second approach you train a model and test it on the same data, right ? This approach overestimates
every performance criterion.
Shortly said: Only compare different models via performance criteria generated by applying X-Validation