Github user imatiach-msft commented on the issue:
https://github.com/apache/spark/pull/22087
@srowen @mgaido91 @erikerlandson I don't think it makes sense to use
PredictionModel as it is currently because it is for supervised learning and
contains label column specific params (eg getLabelCol() method). KMeans and
friends are unsupervised learning and should not have these params. I
definitely agree though that some of the goodness on prediction model should
live on a common base class shared by the supervised and unsupervised algos.
Maybe a class hierarchy like:
predictionmodel
unsupervisedmodel supervisedmodel
would be ideal? where the label specific params would live on
supervisedmodel.
also tagging @jkbradley
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
