"PCA SEVERE: java.lang.ArrayIndexOutOfBoundsException"
maccten
New Altair Community Member
Hi,
I am using rapid miner to try and predict several numeric labels based on numeric attributes.
The dataset has a high deminsionality with many attributes
The attributes are highly correalated so im using a PCA operator to create uncorrealated components. The result of which are fed into a Regression operator to see if it is more effective than using just the regression operator on its own
The issue arises when i change the label. For the example below, the PCA operator works perfectly well when setting the role of Attr1 to label
However when i change the label to Attr2 i get an error SEVERE: java.lang.ArrayIndexOutOfBoundsException
Role Name Type Stats Range Missing Values
regular Attr1 real avg = 69.584 +/- 20.640 [0.000 ; 100.000] 0.0
regular Attr2 real avg = 63.434 +/- 18.475 [0.000 ; 97.143] 0.0
Im new to rapidminer so any help would be greatly appreciated
Thanks
I am using rapid miner to try and predict several numeric labels based on numeric attributes.
The dataset has a high deminsionality with many attributes
The attributes are highly correalated so im using a PCA operator to create uncorrealated components. The result of which are fed into a Regression operator to see if it is more effective than using just the regression operator on its own
The issue arises when i change the label. For the example below, the PCA operator works perfectly well when setting the role of Attr1 to label
However when i change the label to Attr2 i get an error SEVERE: java.lang.ArrayIndexOutOfBoundsException
Role Name Type Stats Range Missing Values
regular Attr1 real avg = 69.584 +/- 20.640 [0.000 ; 100.000] 0.0
regular Attr2 real avg = 63.434 +/- 18.475 [0.000 ; 97.143] 0.0
Im new to rapidminer so any help would be greatly appreciated
Thanks
0
Answers
-
Hi,
can you please post your process setup as described here? If your data is not confidential, it would also be helpful if you could create a bug at http://bugs.rapid-i.com and attach the data (beware, the data will be publicly downloadable from the internet). If the data is too confidential for the internet but public enough for a mail, you could mail me the data for investigation. In that case I can send you my mail address via forum PM.
Best, Marius0