Lightgbm ValidationIndicatorCol - which are values exactly in this str Col

[whiteneverdie]

Hi, folks!

validationIndicatorCol (str): Indicates whether the row is for training or validation

def setValidationIndicatorCol(value: String): this.type = set(validationIndicatorCol, value) }

Is it means just string col with two values "training" or "validation" litteraly?

Originally posted by @whiteneverdie in https://github.com/Azure/mmlspark/issues/689#issuecomment-644167974

[welcome[bot]]

👋 Thanks for opening your first issue here! If you're reporting a 🐞 bug, please make sure you include steps to reproduce it.

[imatiach-msft]

@whiteneverdie great question, and sorry about the confusion. Yes, this should just be the name of the column: https://github.com/Azure/mmlspark/blob/master/src/main/scala/com/microsoft/ml/spark/core/contracts/Params.scala#L179

The column itself should just contain booleans, we filter it here: https://github.com/Azure/mmlspark/blob/master/src/main/scala/com/microsoft/ml/spark/lightgbm/LightGBMBase.scala#L175

Note there is an issue with having huge validation sets, hopefully that's not the case for your scenario: https://github.com/Azure/mmlspark/issues/689

Oh, actually, it looks like you commented on that issue already, sorry I didn't notice your comment there.

[shuDaoNan9]

booleans value in ValidationIndicatorCol, emm. But where can we set what kind of metric for our validation? @imatiach-msft thanks!

[imatiach-msft]

@JWenBin the metric for validation can be set here, using setMetric: https://github.com/Azure/mmlspark/blob/master/src/main/scala/com/microsoft/ml/spark/lightgbm/LightGBMParams.scala#L310

[ICDI0906]

Hi~ I have a question.Why the validate data participate in model training when using setValidationIndicatorCol，Looking forward to your reply.

[ttpro1995]

So, i think in the code should describe more clearly, such as:

• The param contain column name
• That column type is boolean: true if that row is validate set, false if that row is training set