Data Talk S1E1: K-fold Cross Validation with Watson Natural Language Classifier

In this video, join Stefan van der Stockt and Tim Durniak to discuss a project applying k-fold cross-validation. This process employs Receiver Operating Characteristics (ROC) analysis with precision & recall metrics to automate the training for an ensemble of classifiers. Note: This methodology can be used with Watson Natural Language Classifier, Watson Assistant Intents, and Watson Visual Recognition!

Tim Durniak, a Solution Architect and Stefan van der Stockt, a Data Scientist, both are a part of the IBM Watson Delivery organization.

For more information on Receiver Operating Characteristics, see Tom Fawcett’s paper “An Introduction to ROC Analysis”:

Theme Music:
Matt’s Blues by Kevin MacLeod is licensed under a Creative Commons Attribution license (

(Source: IBM Watson,