"X-Validation @ Vector Linear Regression?"

horstmaenhorstmaen Member Posts: 6 Contributor II
edited June 2019 in Help
Hello together,

how can I do an X-Validation for the Vector Linear Regression with multiple labels?
I always get an error message when trying to implement an X-Validation. Without the validation, the operator works just fine.
Or is it not possible for multiple labels using this operator?

Thanks!

(PS: the data consists of polynomial & numeric attributes and a numeric set of labels)
Tagged:

Answers

  • frasfras Member Posts: 93 Contributor II
    Please provide a sample of your data/process otherwise it is hard to imagine what you have done.
  • horstmaenhorstmaen Member Posts: 6 Contributor II
    Hi, thanks for your reply!

    The data consists of 33 Input Variables (partly polynomial and partly numeric) and a set of 32 labels (numeric; sound pressure levels measured for different frequencies). The data set describes (geometry) parameters of gears in a wind power plant and the noise generated by the plant. The aim is to find out which geometry parameters influence sound pressure level the most.
    Unfortunately, I cannot provide actual data, since it's confidential.

    Here's a link of a picture of the process i set up and the error discription:

    http://cdn.imghack.se/images/cf513f000ff68b3497fd83a81112dc36.jpg

    The exact same process worked for other operators such as "Linear Regression". This process also works for "Vector Linear Regression" if I do not use an X-Validation. So my question is whether X-Validation does maybe not work for "Vector Linear Regression" at all due to the set of labels maybe or sth?
    (I've set the labels properly as "label", "label_1", "label_2" etc. using the "set role"-operator)

    Thanks for your help!!!

  • horstmaenhorstmaen Member Posts: 6 Contributor II
    I just need to know if X-Validation is possible with Multiple-Labels at all and how to implement it.
    With a single label I do know how to do it and it works fine.

    Thanks
Sign In or Register to comment.