Trip mode detection from massive smartphone data

Abstract

Nowadays, some smartphone applications require the location of users to be able to provide circumstantial information. However, this data may not be fluid and continuously recorded in a way that can be easily analysed for transport planning purposes. This paper proposes a methodology to reconstruct trips and detect modes from a weather smartphone app data, combined with a validation survey. These results can be useful to create origin-destination matrices and other analyses based on trip data. Our study shows that the Artificial Neural Network (ANN), combined with a proposed data processing framework, provides the best travel mode detection.