Stefan Th Gries
tlooto Summary
A data set is showcased that highlights where such methods can fail at providing optimal results and then solutions to this problem as well as the interpretation of random forests more generally are discussed.
Abstract
Abstract This paper is a discussion of methodological problems that (can) arise in the analysis of multifactorial data analyzed with tree-based or forest-based classifiers in (corpus) linguistics. I showcase a data set that highlights where such methods can fail at providing optimal results and then discuss solutions to this problem as well as the interpretation of random forests more generally.
Citation format
GRIES, Stefan Th. On classification trees and random forests in corpus linguistics: Some words of caution and suggestions for improvement. Corpus Linguistics and Linguistic Theory, 2019, 16: 617–647.