Neural network techniques land use/transport models. Transport systems as well as land-use systems have an effect on spatial development. Land-use influences the demand of (the quality of) infrastructure and vice versa. Subsequently, land-use and transport modelling have to be integrated. In this paper the possibilities of Artificial Neural Networks (ANNs) within transport/land-use modelling is researched. The advantages of ANNs are: short development period, robustness, processor time efficient, and fast adaptation to a changing environment. Possible drawback is their 'black-box' feature. Literature research shows that ANNs are potential methods that are capable of good prediction performances. ANNs have a parallel model structure as opposed to traditional sequential models. This advantage results in an omission of uncertain causally relations. However, ANNs usually lack feedback between input and output. Based on these conclusions we propose a conceptual parallel model that uses ANNs with input/output feedback incorporated. Further research has to result in the actual implementation of the proposed model. (Author/publisher)
Abstract