topic Re: Implement DBSCAN Algorithm on SAS Viya in SAS Data Science

Implement DBSCAN Algorithm on SAS Viya

Edoedoedo — Mon, 26 Aug 2019 09:44:30 GMT

Hi,

I need to perform a DBSCAN clustering on a dataset loaded in CAS. As far as I read, there is no DBSCAN algorithm implemented neither in CAS nor in SAS 9.4. The only clustering method I found in CAS is the K-MEANS but that's inappropriate for my application since I'm working on anomaly detection.

Did I miss something? If DBSCAN is indeed not implemented anywhere in SAS/CAS, what would you recommend me to do? Develop it in CASL? Develop it in SAS Base? Develop it in Python with SWAT?

Thank you for any advice.

PS: the DBSCAN implementation should be with high performance, my dataset has a dozen features and some million rows; I tried the sklearn DBSCAN on my machine and it takes forever, I need to use CAS distributed environment I guess.

Re: Implement DBSCAN Algorithm on SAS Viya

JosvanderVelden — Fri, 30 Aug 2019 09:52:35 GMT

SAS Event Stream Processing (aka ESP) implements DBSCAN. See https://documentation.sas.com/?docsetId=espan&docsetTarget=p1bhz495bvwgapn107kast5bprnq.htm&docsetVersion=5.2&locale=en

DBSCAN is also available in SAS In Memory Statistics (aka IMSTAT): https://documentation.sas.com/?docsetId=inmsref&docsetTarget=p08w4hxeqsgklrn14aqs76gs01te.htm&docsetVersion=2.83&locale=en

Do you know ESP or IMSTAT?

Re: Implement DBSCAN Algorithm on SAS Viya

gsvolba — Fri, 21 Feb 2020 11:23:25 GMT

Hi,
in Viya (CAS Server) you also find a list of additional procedures that are relevant for anomaly detection

Robust Principal Component Analysis (RPCA):
Support Vector Data Description (SVDD):
Isolation Forest (PROC FOREST):
Autoencoder (PROC NNET):
PROC SEMISUPLEARN
Also KCLUS Procedure (allows to use Nominal Variables)

The KCLUS procedure uses the k-means algorithm for clustering interval input variables, uses the k-modes algorithm for clustering nominal input variables, and uses k-prototypes algorithm for clustering mixed input that contains both interval and nominal variables.

Best regards,

Gerhard