"How to easily compute Percentage of Predictions PRED(25) from Cross Validation?"

ikhwan

Dear all,

I need to compute a regression performance measure named "percentage of predictions within 25%" or known as PRED(25). Basically, it requires the predicted values from the whole test set we use in the experiment. The whole procedure to compute PRED(25) as follows:

r = | actualValue - predictedValue |
Pred(25) = count the number of (r / actualValue) which are less than 25% and divide the result by the number of instances

I know that there is possibility to extend a Performance node in RapidMiner, but I think it is not easy for beginner.
In RapidMiner, if I use cross validation, I can only get the test set from last iteration. However, to compute PRED(25), I need the whole test set from each iteration. Is there any way to get the whole test set from a cross validation?

Thanks in advance.

Cheers,
Ikhwan