Resumo de Exploring Alternatives For The Construction Of Neural Spatial Interaction Models

Exploring alternatives for the construction of neural spatial interaction modelsThe rapid growth of Brazilian cities, without a previous planning oftheir expansion (including land use and occupation), causes manyinconveniences for the population related to their transportation, asthey must cover longer distances. This asks for an understanding of thecity areas limits, the services currently offered to the community ineach area, and the users of these services, not only in quantitativeterms, but also in terms of spatial distribution. Moreover, theknowledge of the demand evolution in time and its spatial locationallows the evaluation of many planning scenarios for managing thedemand and the supply, and it is possible, for example, to foresee theregions where the demand is going to be concentrated. Other aspectsthat must be evaluated are the origin, destination and number of tripsthat occur in a determined set of tracts, which can be predicted by thespatial interaction models. Therefore, some studies were made with theobjective of evaluating the performance of Spatial Interaction Modelsbased on Artificial Neural Networks (ANNs). It was observed in thesestudies, some difficulty in selecting the neural network configurationthat best models the problem. As in the majority of research that usesArtificial Neural Networks for the construction of that kind of model,the network parameters are randomly chosen and, even if one can obtainsatisfactory results by varying these parameters, the neural net usedmay not be producing the optimal solution. The objective of this workis to evaluate the use of different alternatives, such as the GeneticAlgorithms (GAs) optimization technique and the bootstrappingestimation method, as supporting tools to select Artificial NeuralNetworks configurations applied to Spatial Interaction Models, and toevaluate the spatial distribution of the residual (errors) predictionresults. The research was developed in a Geographic Information System(GIS) and the data used for this application reflects the changes inthe spatial distribution of the demand for education services in aBrazilian medium-sized city (São Carlos, SP) throughout two years. Theresults obtained showed that although neural models are suitable forestimating transportation flows, gravity models are able to producevery good and precise estimates of the future spatial distribution ofthe demand for educational facilities. This is very important for theplanning process aiming at the reduction of displacement costs ofstudents in the future, given the simplicity of the gravity modelstructure and its straightforward implementation

Resumos Relacionados