topic Re: I wanted to do one hot encoding on categorical variables for Machine Learning . How can I do that in Splunk? in Splunk ITSI

I wanted to do one hot encoding on categorical variables for Machine Learning . How can I do that in Splunk?

jcvytla — Tue, 27 Mar 2018 19:45:21 GMT

How to do label encoding on categorical variables in splunk. I'm new to splunk and trying to explore hidden features. Can I also know how to split the fields as done in excel.

Re: I wanted to do one hot encoding on categorical variables for Machine Learning . How can I do that in Splunk?

aljohnson_splun — Tue, 27 Mar 2018 20:00:15 GMT

When using the Machine Learning toolkit - it will actually convert your categorical variables into indicator variables - columns of 0's and 1's automatically, behind the scenes. Pretty nifty! It uses panda's get_dummies to do this.

If you want to do so manually, you can try using eval:

| eval {fieldToEncode} = 1
| fillnull

The way this works: the { } around the field mean to take the field's value, and use that as the name for the field. Then we assign that to 1, and fill in all the blanks with zeros.

Re: I wanted to do one hot encoding on categorical variables for Machine Learning . How can I do that in Splunk?

jcvytla — Tue, 29 Sep 2020 18:43:09 GMT

Hi @aljohnson

If I want to encode Server_name column , should I use
| eval {Server_name} =1 ?

and extension of that question : suppose i have app_id column , which have values like 1234456122.xxxx

Now I want only the numerical part ,and want to remove the xxxx part...how do I do it in splunk

Re: I wanted to do one hot encoding on categorical variables for Machine Learning . How can I do that in Splunk?

aljohnson_splun — Wed, 28 Mar 2018 16:58:37 GMT

Thats one way you could do it - sure. You'd need to add the fillnull command in order to get the zeros in the columns though, too.

For your second question, you can use eval or rex or many other search commands to do that:

http://docs.splunk.com/Documentation/Splunk/7.0.3/SearchReference/Eval
http://docs.splunk.com/Documentation/Splunk/7.0.3/SearchReference/rex
http://docs.splunk.com/Documentation/Splunk/7.0.3/SearchReference/replace

Re: I wanted to do one hot encoding on categorical variables for Machine Learning . How can I do that in Splunk?

aeapen — Fri, 29 Jun 2018 13:38:37 GMT

@aljohnson_splunk, can label encoding done the same way? I want to categorical variables into numbers with some sort of ranking to it. Is there any method for this?

Re: I wanted to do one hot encoding on categorical variables for Machine Learning . How can I do that in Splunk?

aljohnson_splun — Fri, 29 Jun 2018 15:57:12 GMT

You could use a lookup if the values are static and you know the categories before hand. Otherwise I'd think you need to add a custom algorithm - https://docs.splunk.com/Documentation/MLApp/3.2.0/API/Overview

Re: I wanted to do one hot encoding on categorical variables for Machine Learning . How can I do that in Splunk?

aeapen — Fri, 29 Jun 2018 16:36:52 GMT

thank you for the answer.