The team at Software Advice recently published a slide deck outlining 3 techniques for testing the accuracy of predictive models. The 3 techniques are:
- Lift Charts and Decile Tables
- Target Shuffling
- Bootstrap
Depending upon you situation, goals, and dataset; all 3 are worthy tests. I would say bootstrap is the most common of the 3, but it is always good to have extra tools in your data science toolbelt. If you are unfamiliar with any of the techniques, see the slides below for a quick overview. For a more detailed description, here is a blog post detailing the 3 techniques: 3 Ways to Test the Accuracy of Your Predictive Models.
Leave a Reply