Watson Studio

New SPSS Big Data Algorithms

By Armand Ruiz posted Thu October 08, 2015 02:44 AM

  
Last week we announced the new version of IBM SPSS Modeler 17.1 and Analytic Server 2.1. In this new post I would like to showcase the new algorithms that we included in this new release.

Combined with SPSS Analytic Server now we offer some additional distributed algorithms:

  • Random Trees, Linear Support Vector Machines, Generalized Linear, Linear, Tree

  • Transparently leverage Spark, bringing performance & scalability advantages up to 2.5x

  • If Spark is not present, algorithms will fall back to run in Map/Reduce


Models generated from these algorithms can be scored against any data source.

newAlgos

 

In the coming days we will detail more each of them....stay tuned!
#SPSSModeler
#Spark
2 comments
5 views

Permalink

Comments

Mon October 12, 2015 11:00 PM

Hi Richard,

It is a new product (now on version 2.1 already) to connect to distributed data sources, basically HDFS. Analytic Server executes your stream in parallel in the distributed environment, you don't need to do anything extra just use Modeler as you used to. Most of the new algorithms are prepared to scale to big data and work with HDFS using Analytic Server. This is basically the architecture: http://www.ibm.com/developerworks/cn/data/library/bd-1408-biginsights3.0/images/img005.gif

Mon October 12, 2015 07:05 PM

Armand,

Is Analytics Server a new name for Modeler Server or a completely new product seperate from modeler server?