A short review on the application of computational intelligence and machine learning in the bioenvironmental sciences

Abstract

This paper aims to provide a short review on the application of computational intelligence (CI) and machine learning (ML) in the bioenvironmental sciences. To clearly illustrate the current status, we limit our focus to some key approaches, namely fuzzy systems (FSs), artificial neural networks (ANNs) and genetic algorithms (GAs) as well as some ML methods. The trends in the application studies are categorized based on the targets of the model such as animal, fish, plant, soil and water. We give an overview of specific topics in the bioenvironmental sciences on the basis of the review papers on model comparisons in the field. The summary of the modelling approaches with respect to their aim and potential application fields can promote the use of CI and ML in the bioenvironmental sciences.

Multivariate approaches in species distribution modelling: Application to native fish species in Mediterranean Rivers

“…Conversely, the ANN users need to be experienced in computer programming and models' building so its popularity has been traditionally restricted to the field of scientific research (Conallin et al, 2010). Nevertheless, they represent by far the most popular technique among those encompassed in the group of computational intelligence and machine learning in the bioenvironmental sciences (Fukuda & De Baets, 2012). Nevertheless, scientific community and EFA practitioners are constantly innovating and searching for novel and accurate techniques.…”

“…Although MLP ensembles have been profusely used in many scientific research areas such as running flow forecasting (Abrahart et al, 2012) or lung cancer diagnosis (Zhou et al, 2002a) they have received little attention in ecology. Consequently, although the number of published papers involving Artificial Neural Networks showed a steady increment during the last decade (Fukuda & De Baets, 2012), they have been mainly restricted to training single MLPs. As described in the introduction, habitat suitability and species distribution modelling with ensemble approaches have principally employed treebased techniques such as random forests (e.g.…”