Decision tree classification with bounded number of errors

Aline Medeiros Saettler, Eduardo Sany Laber, Felipe de A. Mello Pereira

Abstract: Highlights•We consider the problem of approximating oblivious decision trees.•We study the more general case where at most k classification errors are allowed.•An approximation algorithm based on randomized rounding is proposed.•We obtain the best possible approximation factor unless P=NP<math><mi mathvariant="script" is="true">P</mi><mo is="true">=</mo><mi mathvariant="script" is="true">N</mi><mi mathvariant="script" is="true">P</mi></math>.