A. Japón Sáez, J. Puerto Albandoz, V. Blanco
In this work we present a novel mathematical optimization-based methodology to construct multiclass optimal classification trees with SVM-based splits. In our approach, the labels of the training observations are temporarily left out in the branch nodes and observations are grouped into two classes which are separated by means of a hyperplane. We provide a Mixed Integer Non Linear Programming formulation for the problem and report the results of an extensive battery of computational experiments.
Keywords: SVM, Optimal Classification Trees
Scheduled
GT01 Localización V. Applications
June 7, 2022 4:50 PM
A15