New advances in spatial network modelling: towards evolutionary algorithms

This paper discusses analytical advances in evolutionary methods with a view towards their possible applications in the space-economy. For this purpose, we present a brief overview and illustration of models actually available in the spatial sciences which attempt to map the complex patterns of spatial/social networks. Particular attention is given to new emerging tools, such as neural networks (NNs) and evolutionary algorithms (EAs), belonging to the field of neurocomputing. The main part of the paper focuses specifically on EAs, in order to better comprehend the potentiality and applicability of this new methodological tool, also in comparison with `conventional' approaches (like logit and spatial interaction models). A historical review of EAs, together with an exposition of their theoretical structure, is offered. Moreover, an empirical application related to the modal split analysis in the European freight transport network shows the potential of EAs in this particular context of analysis. The final section concludes and points to new research directions. It is mainly based on the idea that, since EAs are optimizing techniques generating mechanisms of natural selection and genetics, it is methodologically interesting to investigate their compatibility with the rational behavioural paradigm underlying conventional logit and spatial interaction models (and thus also the potential compatibility between natural selection and economic choice).