Author

Marc-Edouard Schultheiss

Acknowledgements

Links

coming soon

In 2019, the Swiss Digital Strategies integrated a clear agenda about the future of mobilities [2]. Firstly, to foster and accompany the development of new technologies and services; and secondly, to higher the threshold of standard open data. Within the same time frame, business-driven stakeholders (e.g., transport operators, telecom companies) started to offer insights on their data, such as the Urban Mobility platform from Swisscom [3], Movement platform from Uber [4], Metro from Strava [5], or the Tomtom traffic index. Today, geolocation is everywhere. Yet, the generated data is extensive and heterogenous (when accessible). Datasets are large, multi-sourced, often noisy, and come in various formats and standards. In addition, a specificity of urban data is its mix levels of restriction. While geolocation data is private and sensitive, public transit schedules are open data.

In this context, there is a need

This project proposes to develop and release an open Python package to address these four needs and therefore contribute to Urban Mobility Open Research Data practices. This package will complement, enhance or piggyback on a set of existing open python packages [6-11]:

The preprocessing part is probably the most important: how to transform raw geolocation data – basically timestamped longitude/latitude tuples – into mobility data (e.g. trips, modes, point-of-interest) ? Raw geolocation data is valuable because it can be collected seamlessly and widely. It can therefore be representative of the population at low cost and it is not subject to the respondents’ fatigue. However, geolocation data is far from being usable in behavioral analyses as it only contains noisy georeferenced timestamps. And it requires both an extensive expertise in transportation engineering and advanced skills in data science. Today, only a few companies offer to transform raw geolocation data into mobility data (e.g. [12]). But to the best of our knowledge, no open-source library allows to do it in a sufficiently accurate manner.

In addition to transforming raw geolocation data into mobility data, the preprocessing steps include systematic spatial aggregation to purposedly alter the quality of the data (mainly for storage purposes). This can be leveraged as a privacy-controller to choose (a) different levels of aggregation depending on the needs for the final application (e.g., traffic analysis zone, neighborhood, zip code, town); and (b) the geofence the raw data (e.g., remove points in the vicinity of home or work). Put together, this pre-processing ensures that privacy is taken into account from the very first stage of data transformation. This constitutes the building block to comply with the GDPR (General Data Protection Regulation) or the DPA (Data Privacy Act).

Lastly, additional efforts are expected to be conducted for “contextual data” integration. Based on a promising study that leverages transit data to assess the demand-supply gaps in public transport [13], the goal is to integrate General Transit Feed Standard (GTFS) [14] or the Open Street Map (OSM) [15] data. A growing (and well organized) community is contributing to such efforts (see [16]). This data integration must happen at this stage to ensure the correspondence in spatial or coordinate reference systems.

To conclude, x y t abstracts activity space and activity organization as simple standard objects to facilitate scalability and reproducibility. x y t allows privacy-by-design analytics, and a smooth the integration of open contextual data. x y t brings different fields and methods together, at the service of human mobility. x y t covers multiple needs and tend to comply with the emerging practices and skills of researchers and tomorrow’s practitioners. x y t brings a contribution to the transportation-related open research community that gathers more than a hundred contributors: awesome-transit [16], Urban Data Lab [17] and MATSim [11] communities.

Up next →