Don't be so One-Dimensional: How to Engineer Multi-Dimensional High Cardinality Categorical Inputs for Machine Learning

Materials from a paper/talk for Southeast SAS User Group Conference, held on October 23-25, 2022 in Mobile, AL.

Materials provided:

• Full conference paper

• Slides used in presentation

• Two Demonstrations:

  • Demonstration 1 - An implementation of calculating target representation using k-fold cross validation from Section 5 of the paper in .sas format and Python (coming soon).
  • Demonstration 2 - Model trianing and evaluation from Appendix B of the paper in .sas format and Python (coming soon).

Prerequisites

The .sas example file for Demonstration 1 should work in any relatively recent installation of SAS® 9.4 with SAS/STAT. The .sas example file for Demonstration 2 will only work in SAS Viya™ versions 3.5 and above.

Authors

• nikolicxa

Disclaimer

This project is in no way affiliated with SAS Institute Inc.