SAS® Enterprise Miner recently got a major release – 13.1 – focused on machine learning, scalability and productivity. It’s been a while since I blogged about SAS Enterprise Miner (last review here) so this might not be a complete list of the improvements since then.
The machine learning focus added High Performance Support Vector Machines and Clustering while upgrading the HP Neural Network and Random Forest algorithms. From a scalability perspective the Principal Components, GLM, Bayeisan and Time Series Data Mining algorithms were all updated for high performance (multi-threaded parallelism etc). The high performance algorithms work the same and are dropped into SAS Enterprise Miner as nodes the same as always. These high performance data mining algorithms have been added in each of the last few versions. The intent of all this work is to support modelers so they can use all the data they have (in Hadoop, Pivotal, Greenplum, Teradata, and SAP HANA ); model against this data; and do so quickly and iteratively thanks to multi-threaded/parallel processing “HP” or “High Performance” algorithms. SAS Enterprise Miner now has 13 analytical modeling nodes that are “HP” nodes.
Productivity improvements include integration with R as well as support for more integration across the model lifecycle. 13.1 allows R code to be integrated into SAS Enterprise Miner using an R node. These nodes can use any R code. Further, if the R you use can generate PMML, SAS Enterprise Miner can convert the result back to SAS scoring code. When using the R node the SAS Enterprise Miner dataset is mapped to variables to be passed to the R package. The node is a code node so checking, syntax etc is handled at compile time.
Users can also now use a new node in SAS the Enterprise Miner flow to register models with the SAS Metadata Server. Once so registered it can be accessed by SAS Model Manager, SAS Enterprise Guide etc. Similarly another new node allows the data to be saved as a SAS data set, JMP table or spreadsheet for further analysis or manipulation.
You can get more information on SAS Enterprise Miner here and SAS is one of the vendors in our Decision Management Systems Platform Technologies Report.
Comments on this entry are closed.