Machine learning-based framework for data reviewing of a national household travel survey

Abstract

Machine learning techniques have mainly been applied to physical measurement data in the past. In this paper, machine learning is applied to survey data of everyday travel behavior provided by the German Mobility Panel (MOP). The presented model framework supports trained staff in checking trips collected in a trip diary. To this aim, four algorithms are applied and tested further. The neural network (NN) shows the most appropriate results. By using the NN for the individual trip checks, the time effort for the trained staff can be reduced by 20.4 %. In addition, it decreases the number of data samples where all reported trips must be checked. Our study shows that machine learning can support the process of data checking in the MOP leading to significant time reduction.