Github user mengxr commented on the pull request:
https://github.com/apache/spark/pull/3062#issuecomment-72540267
@srowen @selvinsource From user perspective, this is the code they need to
export a model to PMML:
~~~
import org.apache.spark.mllib.export.ModelExporter
val kmeansModel = ...
ModelExporter.toPMML(kmeansModel, "/path")
~~~
They need to remember two things: 1) `ModelExporter`'s name and its
package, 2) whether a model is exportable to PMML or not because `toPMML` takes
`Any`. As a user, I would prefer the following
~~~
val kmeansModel = ...
val pmml = kmeansModel.toPMML
~~~
We can add a trait called `PMMLExportable`, which contains `toPMML`. (We
need to discuss the return type of `toPMML`.) So users know whether a model can
be exported to PMML or not. For the implementation, I agree with @srowen that
`.pmml` package should be sufficient because there is no other exchangeable
formats for predictive models. The real export code for each model should live
under this package as private and `toPMML` in each model becomes a simple
wrapper.
Due to lacking feature attributes, I'd like to mark everything as private
for now until we add ML attributes.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
