
Set up Predictive Analytics in ITSI
Before you can predict service health scores with ITSI Predictive Analytics, you need to configure the Splunk Machine Learning Toolkit application and optionally set up permissions.
Enable Predictive Analytics
To use Predictive Analytics, you need to install the Splunk Machine Learning Toolkit (MLTK) and share the machine learning macros with all apps so ITSI can access them. To minimize memory and disk consumption, configure the Machine Learning Toolkit configuration file. This will allow the MLTK to handle more memory and events.
- Install the Python for Scientific Computing add-on version 1.3 or later for your operating system:
- Install the Splunk MLTK version 3.2.0 or later. Follow the steps in Install the Splunk Machine Learning Toolkit in the MLTK User Guide.
- Give the MLTK app Global permissions:
- In ITSI, click App: IT Service Intelligence > Manage Apps.
- In the filter bar, enter
Splunk Machine Learning Toolkit
. - Click Permissions.
- Ensure that All apps is selected.
- Click Save.
- (Optional) Configure the MLTK to handle more memory and events:
- Create a copy of
mlspl.conf
in$SPLUNK_HOME/etc/apps/Splunk_ML_Toolkit/local
. - Configure the following settings under the [default] stanza and in each algorithm-specific stanza:
Setting Recommended value max_inputs
1000000 max_memory_usage_mb
2000 max_model_size_mb
30 max_fit_time
2400 - Save the file and restart Splunk.
- Create a copy of
Permissions requirements
The following table summarizes the ITSI roles and corresponding Predictive Analytics capabilities:
Role | Capabilities |
---|---|
itoa_admin |
|
itoa_team_admin |
|
itoa_analyst |
|
itoa_user |
|
For more information about ITSI role capabilities, see Configure users and roles in ITSI in this manual.
PREVIOUS Predict and prevent outages with ITSI Predictive Analytics |
NEXT Access Predictive Analytics in ITSI |
This documentation applies to the following versions of Splunk® IT Service Intelligence: 4.0.0, 4.0.1, 4.0.2, 4.0.3, 4.0.4, 4.1.0, 4.1.1, 4.1.2, 4.1.5
Feedback submitted, thanks!