All Apps and Add-ons

MLTK training run failing with 'Usecols do not match columns, columns expected but not found' error

dfphere
Explorer

I'm utilizing Principal Component Analysis (PCA) on a RandomForestRegressor model to process some of the text fields in my data, which results in a certain number of PCA fields (around 30, I would say). The model look good upon the initial `fit` from within the experiment window, so I saved the model and scheduled a training run to occur every morning.

However, the scheduled training fails with a 'Usecols do not match columns, columns expected but not found' failure. It normally reports a handful of PC_* fields on the higher end of the range (like PC_27 - PC_31) not being found. The error appears to be related directly to the pandas python library but I don't have the capability to troubleshoot the code itself and hoping to resolve the issue via MLTK. Can anyone assist?

Labels (2)
Tags (3)
0 Karma
Get Updates on the Splunk Community!

AppDynamics Summer Webinars

This summer, our mighty AppDynamics team is cooking up some delicious content on YouTube Live to satiate your ...

SOCin’ it to you at Splunk University

Splunk University is expanding its instructor-led learning portfolio with dedicated Security tracks at .conf25 ...

Credit Card Data Protection & PCI Compliance with Splunk Edge Processor

Organizations handling credit card transactions know that PCI DSS compliance is both critical and complex. The ...