D. Conde del Río, M. Fernández Temprano, C. Rueda Sabater, B. Salvador González
The problem of classifying observations in one of k classes is receiving a lot of attention due to its applicability in a wide range of problems, and many methods and techniques to build classification rules have been developed. Among these, boosting procedures, based on weak classifiers, have been recently developed. However, none of the standard procedures are able to take advantage of the existence of additional information, frequently appearing in applications, that can be expressed in terms of order restrictions. Here, we define new procedures that allow us to incorporate monotonicity constraints to boosting and backfitting procedures in the additive logistic model.
We show the good performance of the new procedures, comparing them with several well-known classification rules, both in a simulation study and applying them to publicly available data sets with monotonicity constraints on some predictors and an ordinal response variable.
Palabras clave / Keywords: boosting, backfitting, monotonicity constraints, classification
Programado
Sesión GT02-2: Análisis Multivariante y Clasificación (AMyC-2). Organizadora: Eva Boj del Val
29 de mayo de 2018 12:20
Sala 5
