Skip to content

Boosted regression trees

2 messages · twilding, Sam Veloz

#
This paper came out recently - a guide to boosted regression trees for 
ecologists. Great potential for modelling species distributions from 
environmental data I think.

http://www.blackwell-synergy.com/doi/abs/10.1111/j.1365-2656.2008.01390.x

Well written (even I can understand it) and code for R is given with 
tutorials. Example for freshwater fish given. The method's non-linear. It 
doesn't produce nice smooth curves, but that seems to be it's strength as no 
prior assumptions are made about how the species should respond. I haven't 
tried it yet. Interested to hear what sort of results people get.

Thomas Wilding
Department of Biology
Colorado State University
Fort Collins
Colorado 80523-1878
USA
(+1 970) 491 2414
#
The code they provide is really easy to use and so far I have gotten 
great results. Does require a little of work though because you do have 
optimize several parameters, which they discuss in the paper. Also, 
because you are often running >1000 iterations, running each model takes 
much longer than a glm for example. So far with my work I have found 
better predictive accuracy using boosted regression trees vs GLM's or 
multi adaptive regression splines.

Sam
twilding wrote: