Neural network and spatial interaction analysis of European commodity flows

The present paper aims to analyse interregional freight transport movements in Europe in order to forecast spatio-temporal patterns of new transport economic scenarios. In view of the high dimension of our data-base on transport flows, two different approaches are compared, viz. the logit model and the neural network model. Logit models are well-known in the literature; however, applications of logit analysis to large samples are more rare. Neural networks are nowadays receiving a considerable attention as a new approach that is able to capture major patterns of flows, on the basis of fuzzy and incomplete information. In this context an assessment of this method on the basis of a large amount of data is an interesting research endeavour. The paper will essentially deal with a research experiment, oriented towards both calibration/learning procedures and forecasting, in order to compare the two above methodologies as well as to investigate the potential/limitations of the two above mentioned different, but related assessment methods. The first results in this framework highlight the fact that the two models adopted, although methodologically different, are both able to provide a reasonable spatial mapping of the interregional transport flows under consideration.