Is it still recommended to use spark-ts?

129 views
Skip to first unread message

lppier

unread,
Mar 3, 2019, 8:33:44 PM3/3/19
to Time Series for Spark (the spark-ts package)
Hi, 

I noticed that development on this library ceased sometime around 2016. 
If I am trying to use Spark for time series prediction, what is the de-facto library as of now? 
Or, am I missing something and it is counter-intuitive to do time-series on Spark?
Thank you. 

Pier. 

Sandy Ryza

unread,
Mar 3, 2019, 11:24:16 PM3/3/19
to lppier, Time Series for Spark (the spark-ts package)
Hey Pier,

That's correct - this library is no longer under active development and has lots of gaping holes.

It looks like Flint (https://github.com/twosigma/flint/commits/master) has had some more recent development.

It also depends what you mean by "do time series".  If you're looking for time series statistical models, and have time series that can fit on a single machine (which I think is most time series), there's always the option of using statsmodels inside PySpark.

-Sandy

--
You received this message because you are subscribed to the Google Groups "Time Series for Spark (the spark-ts package)" group.
To unsubscribe from this group and stop receiving emails from it, send an email to spark-ts+u...@googlegroups.com.
To post to this group, send email to spar...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/spark-ts/48c7f4c2-53cc-446f-bf1f-4ea5124eef4e%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Pier Lim

unread,
Mar 4, 2019, 5:21:30 AM3/4/19
to Sandy Ryza, Time Series for Spark (the spark-ts package)
Thanks Sandy for the quick response, appreciate it. I’ll look into Pyspark route !

 

From: Sandy Ryza <sand...@gmail.com>
Sent: Monday, March 4, 2019 12:24 PM
To: lppier
Cc: Time Series for Spark (the spark-ts package)
Subject: Re: Is it still recommended to use spark-ts?
 
Reply all
Reply to author
Forward
0 new messages