[
https://issues.apache.org/jira/browse/SPARK-25941?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Sean Owen resolved SPARK-25941.
-------------------------------
Resolution: Not A Problem
This isn't a bug. The implementation has changed in Spark and MLlib a lot over
time. I would not expect exactly the same answer, and this is quite close.
> Random forest score decreased due to updating spark version
> -----------------------------------------------------------
>
> Key: SPARK-25941
> URL: https://issues.apache.org/jira/browse/SPARK-25941
> Project: Spark
> Issue Type: Bug
> Components: Deploy, Input/Output, ML
> Affects Versions: 2.3.2
> Reporter: jack li
> Priority: Major
> Labels: ML, forest, random
>
> h3. Problem description
> I use different versions of spark to analyze random forest scores..
> * spark-core_2.10 and version 2.0.0
> ** RandomForestsKaggle Score = 0.8978765219058574
> * spark-core_2.11 and version 2.4.0
> ** RandomForestsKaggle Score = 0.8886987035251259
> Source : [https://github.com/smartscity/Kaggle_Titanic_spark]
> [Example github source and
> readme|https://github.com/smartscity/Kaggle_Titanic_spark/blob/master/README.md]
>
> h3. Introduce
> This case is Titanic Competitions on the Kaggle.
> [https://www.kaggle.com/c/titanic]
> h3. Conclusion
> After upgrading the spark version({{version 2.4.0}}), the random forest score
> dropped({{0.01}}).
> h3. Expectation
> Expect random forest score not to drop as the version upgrades.
>
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
