Isotonic boosting classification rules


R code developed by:

Programmed by: David Conde, Department of Statistics and Operational Research, Universidad de Valladolid, Valladolid, Spain

Version: 1.0 (July 31, 2019)


In many real classification problems a monotone relation between some predictors and the classes may be assumed when higher (or lower) values of those predictors are related to higher levels of the response. The incorporation of this information into classification rules allows defining rules that are more easily interpretable and outperform those that do not consider the isotonicity relationships. The algorithms proposed in [1] consider isotonic regression (Robertson et al., 1988) to integrate the isotonicity into boosting classification procedures.


The implemented R code contains the isotonic boosting classification algorithms proposed in [1], namely, SILB and MILB for binary classification, ASILB, AMILB, for multiclass classification under the Adjacent categories logit model and CSILB and CMILB for multiclass classification under the Cumulative probabilities logit model (Agresti, 2010). Also enclosed are the scripts used for obtaining the results described in [1] for the induction motor data example.


Software Info:

The zip file below contains a documented file with the new isotonic boosting procedures, another file with induction motor data and the code to be run to obtain the results related to this example presented in the paper.



[1] Conde, D., Fernández, M.A., Rueda, C. and Salvador, B. (2020). Isotonic boosting classification rules. Advances in Data Analysis and Classification,


Isotonic-Boosting-R-Code, version 1.0 (124 KB)




David Conde, Ph.D.                    
Department of Statistics and OR

Paseo de Belén, 7 
47011, Valladolid 
Tel 983-18-5877