Splunk Machine Learning Toolkit

This image isn’t mine, all rights reserved by Splunk

How does the SMLT learn?

In the Batch Learning Version, there is a 30 day rolling window to predict One Day in Real Time. Following this paradigm, there is no retain of any knowledge from before the 30 day window.

An alternative is to use the Partial Fit, in which there is no nead to reload data after 30 day, we use the same.

Cross Validation in SMLK

| inputlookup dataset.csv | fit SVM = C=0.1 "class_to_predict" from <class1> <class2> ...  <class n> kfold_cv=5

This applies KFold Cross Validation

Another way to do it is to do it manually:

 | inputlookup dataset.csv | sample partitions=5 seed=1| where partition_number!=0 | fit ... into your_model |