Register the custom algorithm in the MLTK app

To use the custom algorithm in Splunk Machine Learning Toolkit and for it to be visible in the Splunk platform, you must register the algorithm in the MLTK app. You can register the name of an algorithm using either a manual file update or using the REST API.

To register an algorithm, update the algos.conf file with the name of the custom algorithm you want to add. 

$SPLUNK_HOME/etc/apps/Splunk_ML_Toolkit/local/algos.conf

Register by manual file update

Use the following steps to register the name of an algorithm by manual file update:

  1. Create or update algos.conf in the following directory:
    $SPLUNK_HOME/etc/apps/Splunk_ML_Toolkit/local
  2. Add the algorithm name as a stanza, meaning within brackets, to the algos.conf file: 
    [<Your new algorithm class name>]
  3. Select Save.
  4. Restart Splunk Enterprise.

Register using the REST API

Note: You must have administrator permissions in your Splunk platform instance to use this registration method.

Use the following code to register the name of a custom algorithm with the REST API: 

$ curl -k -u admin:<admin pass> https://localhost:8089/servicesNS/nobody/Splunk_ML_Toolkit/configs/conf-algos -d name="<Your new algorithm class name>"

Implement the algorithm

To implement the algorithm, create and name a python script file (.py file) for the algorithm in the following directory: 

$SPLUNK_HOME/etc/apps/Splunk_ML_Toolkit/bin/algos

The name of the .py file and the name of the main class in the .py file must be the same as the stanza name, the name in brackets, used in the algos.conf file. For example, [LinearRegression] for the LinearRegression algorithm. 

<Your new algorithm class name>.py

Note: The name of the algorithm class must be unique and not conflict with other names in the algos.conf file.