OD Mobility Estimation Using Artificial Neural Networks

Abstract

Origin-destination (OD) matrix estimation is an important field in urban and transportation planning frameworks. This matrix gives information on the transportation made between different points of an area. This information is contained in a target OD matrix, which is a result of the data collection phase. The information comes from a sample survey. The purpose of this study is to analyze the mobility in a city performing methods based on soft computing. In order to avoid underfitting due to the high dimensionality of the problem a feature selection stage has been applied and the estimation of the OD matrix was carried out using a hybrid approach based on Artificial Neural Networks (ANNs). The methodology consists of a resampling strategy with crossvalidation using different quality indexes to compare the different approaches in order to achieve the generalization capabilities of the best models. The results obtained prove that the proposed approach could be an effective tool to estimate OD mobility.