[
https://issues.apache.org/jira/browse/SPARK-7338?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15228731#comment-15228731
]
Anahita Bhiwandiwalla commented on SPARK-7338:
----------------------------------------------
I am currently working on implementing this in Spark. Implementing it using the
log of the partial likelihood of the maximum likelihood estimate. My
implementation, as part of this story, is when we have unique survival times. I
am working on having efficient passes on data to not compromise performance. I
have kept the API similar to AFT. This work should be completed in about 3-4
weeks, with testing.
Subsequently, after completing this, for ties in survival times, I am looking
at both Efron approximation(default in R) and and Breslow approximation(default
in SAS). We could keep this as a option- the user can pick which to have chose
to have and either as a default. This could be opened as a new
feature/enhancements ticket.
Please assign this ticket to me.
> Survival Modelling - Cox proportional hazards
> ---------------------------------------------
>
> Key: SPARK-7338
> URL: https://issues.apache.org/jira/browse/SPARK-7338
> Project: Spark
> Issue Type: New Feature
> Components: MLlib
> Reporter: Amit Gupta
>
> One very important model in space of Survival Modelling is missing from Spark
> MLLib v1.3.1. It would be good if Cox Proportional Hazards could be available
> in next release.
> Please refer to link for more details:
> http://en.wikipedia.org/wiki/Proportional_hazards_model
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
