on 08-22-2016 4:47 PM
Hello everyone,
I'm trying to create a row filter in the predict panel of expert analytics to exclude missing values. If I filter under the prepare panel those values are stated as <<null>>. Im thinking of something like:
[Variable] != << null >>
But this command gives an error message. What is the internal representation of missing values in PA?
Thank you very much
Christoph
Hi,
1. You could use R Extension to do data imputation... one possible library that you could use is rfImpute.
2. Depending on which algorithm you are using , they do have configuration parameter called missing value handling, which helps the user decide the action that he wants if a missing value is found in the data. I have shared a screenshot of one algorithm here...
Hope this helps
Thanks
Paul
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
Hi Christoph,
It would be nice to post this idea on the idea place SAP BusinessObjects Predictive Analytics: Home . Though we have this is in the backlog, the votes on this idea on ideaplace will help push this high on the priority list.
Best
Paul
Hey Paul,
I agree. Writing an R extension is a good idea but it might lead to performance issues when it comes to big data. Then again the issue isnt so big if the data source is HANA. Yesterday I wrote a litte R extension to row filter the data set for missing values in multiple columns, returning either the complete or the missing cases. This might help for starting data inspection and further data preparation. If someone is interested im happy to share the script.
Best
Christoph
Hi Christoph, why not making this part of the SAP Analytics partner extensions? I do not see such extension here: https://analytics-extensions.enter.sap/sap/aed/ui/?product=predictive
Could you please be so kind to make the thread "Answered"?
Thanks & regards
Antoine
Hi everyone,
we just published an extension on filtering out missing values in the SAP Analytics Extensions Directory, a detailed descripiton of which you find on our website.
Best
Christoph
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
Hello Christoph, can you please add a screenshot of what you tried to do?
Thanks & regards,
Antoine
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
User | Count |
---|---|
84 | |
10 | |
10 | |
9 | |
7 | |
7 | |
6 | |
5 | |
4 | |
4 |
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.