Abstract
Machine learning is a growing computational field that borrows concepts
and methodologies from statistics and data science to create
semi-autonomous programmes capable of adapting to a multitude of
problems and decision-making scenarios. With its potential in big data
analysis, machine learning is particularly useful for tackling global
conservation problems that often involve vast amounts of data and
complex interactions between variables. In this systematic review, we
summarise the use of machine learning methods in the study of species
threats and conservation measures, and their emergent trends. Maximum
entropy, Bayesian and ensemble methods have gained wide popularity in
the past years and are now commonly used for multiple problems. Their
relevance to modern conservation issues (and associated data types),
their relatively simple implementation, and availability in a variety of
software packages are the most likely factors to explain their
popularity. Neural networks, decision trees, support-vector machines and
evolutionary algorithms have been used in more specific situations, with
some applications showing promise in dealing with increasingly complex
data and
scenarios.