Building Predictive Models with Rule Lists

Click here to enter the CORELS website


Rule lists are fully human-interpretable, giving them distinct advantages over black box models.


Our algorithms utilize highly optimized vector operations, allowing them to run in reasonable time on commodity laptops.


On many datasets, rule lists have been shown to be comparable in accuracy to much more complex black box models.


All our code is free, open source, and under the GNU General Public License v3.0. You can find it on GitHub.

Website by Vassilios Kaxiras and Aditya Saligrama