Boosting, Bagging and Ensembles in the Real World: An Overview, some Explanations and a Practical Synthesis for Holistic Global Wildlife Conservation Applications Based on Machine Learning with Decision Trees

Abstract

Boosting, bagging and ensembles are intellectually `deep' modeling methods well-known and described for several decades. Great computing tools exist to use those methods. But with few exceptions they have not been used well for natural resource conservation management or ecology; for instance, the advanced works of Breiman (2001), Friedman (2001), and Elder (2003) still await generic recognition. Here I present on these methods, conveniently driven by binary recursive partitioning (Classification and Regression Trees CARTs), and many of their real-world aspects and usages. I elaborate on applications and on some of the implementation hurdles known. It is shown that those machine learning methods are the essential part of the new generation of quantitative reasoning. It allows for relevant progress, all while the global environmental state decays further, climate change remain unaccounted for and sustainability policies remain outdated urging for an effective change of global culture and governance.