search

jwijffels / RMOA

Connect R to MOA for massive online data stream mining
37 stars 19 forks source link

Regression MOAOptions being ignored #14

Closed jasvale closed 3 years ago

jasvale commented 3 years ago

Hello,

I'm having an issue with RMOA 1.1.

It seems that regression algorithms are ignoring the passed in options when trying to tune the models.

I've tested with MOA, and the results are as expected, dependently of the parameters, the MSE is different. However this behavior is not seen in RMOA.

For instance:

ctrl <- MOAoptions(model = "FIMTDD", PageHinckleyThreshold = 0.5)
FIMTDD(control=ctrl)

Will print PageHinckleyThreshold as 0.5 in the output. But no changes are seen in the prediction results. I've tried multiple parameters, along with multiple values. Nothing seems to get picked up.

For classification algorithms all seems to be working properly.

I wonder If this is a bug, or something that I am missing?

Thanks in advance.

jasvale commented 3 years ago

It seems I was miss something, my analysis was incomplete in the bigger picture. Please disregard this, it's working.