Optimal Binning for Finding High Risk Cut-offs (1445 Families)

  • Ton J. Cleophas
  • Aeilko H. Zwinderman


Optimal binning is a so-called non-metric method for describing a continuous predictor variable in the form of best fit categories for making predictions. Like binary partitioning (Machine Learning in Medicine Part One, Chap. 7, Binary partitioning, pp 79–86, Springer Heidelberg Germany, 2013) it uses an exact test called the entropy method, which is based on log likelihoods. It may, therefore, produce better statistics than traditional tests. In addition, unnecessary noise due to continuous scaling is deleted, and categories for identifying patients at high risk of particular outcomes can be identified. This chapter is to assess its efficiency in medical research.

Supplementary material

333106_2_En_61_MOESM1_ESM.sav (42 kb)
optimalbinning (SAV 42 kb)

Copyright information

© Springer Nature Switzerland AG 2020

Authors and Affiliations

  • Ton J. Cleophas
    • 1
  • Aeilko H. Zwinderman
    • 2
  1. 1.Department Medicine Albert Schweitzer HospitalDordrechtThe Netherlands
  2. 2.Academic Medical CenterDepartment Biostatistics and EpidemiologyAmsterdamThe Netherlands

Personalised recommendations