A GENETIC-ALGORITHMS BASED EVOLUTIONARY COMPUTATIONAL NEURAL-NETWORK FOR MODELING SPATIAL INTERACTION DATA

Building a feedforward computational neural network model (CNN) involv es two distinct tasks: determination of the network topology and weigh t estimation. The specification of a problem adequate network topology is a key issue and the primary focus of this contribution. Up to now, this issue has been either completely neglected in spatial applicatio n domains, or tackled by search heuristics (see Fischer and Gopal 1994 ). With the view of modelling interactions over geographic space, this paper considers this problem as a global optimization problem and pro poses a novel approach that embeds backpropagation learning into the e volutionary paradigm of genetic algorithms. This is accomplished by in terweaving a genetic search for finding an optimal CNN topology with g radient-based backpropagation learning for determining the network par ameters. Thus, the model builder will be relieved of the burden of ide ntifying appropriate CNN-topologies that will allow a problem to be so lved with simple, but powerful learning mechanisms, such as backpropag ation of gradient descent errors. The approach has been applied to the family of three inputs, single hidden layer, single output feedforwar d CNN models using interregional telecommunication traffic data for Au stria, to illustrate its performance and to evaluate its robustness.