I want to know whether Splunk Machine Learning depends on time series data ? I know predict command totally depends on time series data. But, suppose if the client has logs/data which does not contain timestamps, then in that case, how we can use Splunk ML for predicting failures or impact on services ?
If ML supports only time series data for prediction, then what would be Splunk solution for data without timestamps and how we can predict failures, as well as predict impact on services such as storage/ldap etc.. based on historical data without timestamps ? Please advise.
Good question.. I just happen to be reading up on the ML-Toolkit and was in this section concerning Algorithms in the kit. ML-Toolkit Algorithms .. There is also some more info in the Assistants category on the left side of the page that may provide additional assistance to you while you wait for a real answer.
Most of the Machine Learning Toolkit does not rely on the presence of a _time field. In fact, many of the examples in the Showcase are not time series. You can use the fit and apply commands to predict numeric or categorical fields given other fields. The Predict Numeric and Predict Categorical assistants (especially the examples) should point you in the right direction.