What do mobile network data really tell us about people’s movements? How can we better interpret trips and stops?
The widespread use of connected devices, such as smartphones and vehicle on-board systems, has made available a growing amount of data automatically generated by users. In this context, mobile network data, i.e. information derived from interactions between mobile phones and the mobile network infrastructure, are emerging as a strategic resource for mobility studies.
The FS Research Centre, in collaboration with Motion Analytica, Fastweb and Vodafone proposed a study on this topic, presenting it during the Final Conference of the MNO-MINDS Project promoted by Eurostat “Integrating MNO Data into Official Statistics: challenges and innovations” held at the OECD Auditorium "La Muette" in Paris.
The study begins with a methodological reflection: although mobile data allow for detailed, up-to-date, and large-scale spatiotemporal reconstructions, they were not specifically designed for tracking movements. Raw data do not allow us to distinguish between stationarity and movement, nor do they provide information about the mode or purpose of travel. Therefore, there is a need to apply complex algorithms: data cleaning, trajectory segmentation, transport mode inference, sample expansion techniques and anonymization in accordance with privacy regulations.
In this presentation the research team starts from the definition of a trip as a sequence of “points” separated by “stops”. They explore the crucial issue of using thresholds: a temporal threshold, which sets a minimum duration below which a stop is considered negligible, and a spatial threshold, which helps identify significant stops made by the user during the trip. However, these thresholds can lead to misclassifications because they overlook behavioral differences among individuals.
To overcome these limitations, an alternative data-driven paradigm is proposed, in which the significance of a stop is defined in relation to the individual’s behavior and the actual route taken, compared to the expected route between origin and destination.
This approach allows for a more authentic reconstruction of trips, better reflecting behavioral diversity and individual travel habits. In practice, it improves the accuracy of mobility analysis, reduces the risk of misinterpretation, and provides a more robust foundation for studies and simulations based on mobile data.