[jira] [Commented] (SPARK-19714) Bucketizer Bug Regarding Handling Unbucketed Inputs

["Wojciech Szymanski (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/SPARK-19714?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15881568#comment-15881568 ] 

Wojciech Szymanski commented on SPARK-19714:
--------------------------------------------

IMHO Bucketizer works as expected. I guess that from your point of view, invalid value is a number out of range, i.e. 0,1,2,3,4, but from Spark point of view, invalid value is not a number.
{code}
    if (getHandleInvalid == Bucketizer.SKIP_INVALID) {
      // "skip" NaN option is set, will filter out NaN values in the dataset
      (dataset.na.drop().toDF(), false)
    }
{code}

I fully agree that dosc for handleInvalid might be confusing, since definition of invalid values is missing:
{code}   
    /**
    * Param for how to handle invalid entries. Options are 'skip' (filter out rows with
    * invalid values), 'error' (throw an error), or 'keep' (keep invalid values in a special
    * additional bucket).
    * Default: "error"
    * @group param
     */
    val handleInvalid: Param[String]
{code}

I would suggest that I update the dosc by clarifying what kind of invalid values will be filtered out if 'skip' strategy is used. 
I am not sure if introducing a new strategy for handling values out of range will be welcomed by the community.

> Bucketizer Bug Regarding Handling Unbucketed Inputs
> ---------------------------------------------------
>
>                 Key: SPARK-19714
>                 URL: https://issues.apache.org/jira/browse/SPARK-19714
>             Project: Spark
>          Issue Type: Bug
>          Components: ML, MLlib
>    Affects Versions: 2.1.0
>            Reporter: Bill Chambers
>
> {code}
> contDF = spark.range(500).selectExpr("cast(id as double) as id")
> import org.apache.spark.ml.feature.Bucketizer
> val splits = Array(5.0, 10.0, 250.0, 500.0)
> val bucketer = new Bucketizer()
>   .setSplits(splits)
>   .setInputCol("id")
>   .setHandleInvalid("skip")
> bucketer.transform(contDF).show()
> {code}
> You would expect that this would handle the invalid buckets. However it fails
> {code}
> Caused by: org.apache.spark.SparkException: Feature value 0.0 out of Bucketizer bounds [5.0, 500.0].  Check your features, or loosen the lower/upper bound constraints.
> {code} 
> It seems strange that handleInvalud doesn't actually handleInvalid inputs.
> Thoughts anyone?



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)

---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@spark.apache.org
For additional commands, e-mail: issues-help@spark.apache.org