Empirical models in ecology: some applications of artificial neural networks (1998). 

Stazione Zoologica "A. Dohrn" di Napoli, Villa Comunale, 80121 Napoli, Italy
 
 

Artificial neural networks have been recently introduced in the ecological modeling toolbox. They are especially suited to build up "black box" models that can easily outperform conventional empirical models (e.g. multiple linear regression). Moreover, artificial neural networks can be used to model complex processes even when causal relationships are unknown or not fully understood, since they can to learn empirical rules from large, heterogeneous data sets and summarize them as simple matrix models.

Some applications of artificial neural networks to ecological modeling problems will be presented and their results will be compared with other empirical models. The application examples will show how artificial neural networks can be used to build up models using information sources that conventional models cannot fully exploit (e.g. binary or nominal data, geographical coordinates, etc.).

The role of sensitivity analysis will be also discussed, both as a method for defining the role of each input variable in predicting the "black box" outputs and as a way to optimize the structure of the artificial neural networks. Finally, new potential applications of artificial neural networks to ecological modeling will be outlined.