Decision trees - list of misclassified instances

PedroM

Hello all,

I've just created a classifier based on a decision tree here in RapidMiner, and used X-validation to measure its accuracy.
After running the process a confusion matrix was created, showing a small number of misclassified instances. I would like to know which instances are those (not just the number), but I couldn't find how to do it  >:(
Can someone please help me with this?

Thank you very much!

steffen

Hello Pedro

Are you angry about yourself ?

If you make a breakpoint after the "modelapplier"-operator (select the operator and right-click to see this option), you can see that you both the predicted and the original label available before the confusion matrix is calculated. If this is not enough, you can use tthe operator ...

- "Filter Examples" with the wrong/correct prediction option to filter the erroneous
- use a "Write"-Operator to write them out

hope this was helpful,

steffen

PedroM

Hi Steffen,

That was very helpful, thanks! Your first solution really does the job, but I'd prefer the second one (more elegant) if only I could make it work.
The problem is that when I write the prediction to an xls file, only one of the iterations of the X-validation shows up. The "Write as Text" operator seems to be the only one to append the info of all iterations, but it is hard to read.
Is there any obvious workaround to deal with is? Otherwise, no problem, I will compile the information I already have manually.

Once again, thx for your help,
Pedro
 

IngoRM

Hi,

this could sound like magic right now but just try to add

%{a}

somewhere in the filename you used for writing the XLS file. After that you will end up with k files - one for each iteration. Welcome to the world of RapidMiner macros 

Cheers,
Ingo

PedroM

Cool! It worked out just perfect 

Thanks Ingo and Steffen for your valuable inputs.

Greetings,
Pedro