Ruifeng Zheng created SPARK-58994:
-------------------------------------
Summary: Fix prediction column metadata in GBTRegressionModel
Key: SPARK-58994
URL: https://issues.apache.org/jira/browse/SPARK-58994
Project: Spark
Issue Type: Bug
Components: ML
Affects Versions: 5.0.0
Reporter: Ruifeng Zheng
GBTRegressionModel.transform aliases its prediction expression with featuresCol
and the feature
column metadata. Although withColumns assigns the final prediction column name,
its outer alias
inherits the feature metadata from the inner alias. The transformed prediction
column can therefore
carry vector feature metadata instead of the numeric prediction metadata
produced by
transformSchema.
The fix uses predictionCol and its output metadata for the prediction
expression and adds a
regression test comparing the transformed metadata with transformSchema.
Pull request: https://github.com/apache/spark/pull/58277
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
