Machine Learning Methods without Tears: A Primer for Ecologists

Abstract

There is widespread agreement that machine learning methods—a collection of statistical approaches with AI roots—hold
enormous potential to improve our capacity to comprehend and anticipate environmental occurrences. Ecological system
modeling is a perfect fit for these modeling approaches because of their superior performance compared to more conventional
methods (such as generalized linear models) and their adaptability to complicated situations involving several interacting parts.
Literature reviews show that, in comparison to other fields, ecology makes very little use of these methods, despite their obvious
benefits. The fact that machine learning methods do not fit cleanly into the category of statistical modeling tools that the majority
of ecologists are acquainted with might be one reason for the lack of interest. This study introduces three machine learning
methods that ecologists might utilize in their work: evolutionary computation, artificial neural networks, and classification and
regression trees. We provide a concise overview of the technique, ecological examples of its use, details of model construction
and execution, pros and cons, statistical software availability, and an example for each approac.