Análise dos modelos híbridos na distribuição de viagens : Caso: Medellín, Colômbia

Abstract

In transportation planning, it is important to understand the way that displacement is carried out in a region. Posteriorly, the information will allow to establish a transport policy. For that, it is relevant to known the individual’s behaviors. This research studies the performance of neural networks from the understanding the origins of the journeys, by presented a new approach through the structure of a hybrid model which incorporates aggregate and disaggregate data. In addition, the individual information (socioeconomic) and the dissuasive elements such as time travel and intervening opportunities, will also be used the characteristics around the area (land use) such as location of points of origin, activities dispersal (destinations) and purpose of travel. The results indicates that the travel can be explained by the combination of aggregate and disaggregate variables, and the omission of theses last ones could lead the estimation to high levels of error. On the other hand, it was noticed that each structured model the attraction of destination zone and the impedances are relevant in the travel estimates.