Mobile phone indicators and their relation to socio-economic indicators
Ever wondered what the effect of large scale infrastructure works such as a new railway or a large housing development would have on the way we live and work? In this project we aim at exactly that question. Based on data from census, commuting figures, and timetables from public transport we are developing a model that assesses the nation-wide impacts of large infrastructure works on the UK. It is our so-called Quant model, and we are making it available as a web-based tool so that you too can explore impacts of future changes. So browse here and start building these railways, housing developments or workplaces yourself and see how you change lives!
Collaborators on this project are: Mike Batty, Roberto Murcio, Juste Raimbault, Richard Milton, Natalia Zdanowska, Antonia Godoy-Lorite, Iacopo Iacopini and Elsa Arcaute, all at CASA, UCL.
We are in full development of this piece of research but you can have a taste of what the Quant model can do at this (interactive website): http://quant.casa.ucl.ac.uk/ or you can have a look at my (very short) talk on spatial interaction models at Carto's most recent spatial data science conference here.
Mobile phone indicators and the socio-economic structure of cities.
%20CoverPictur.jpg)
The fact that we can construct behavioral indicators from mobile phone data opens up questions to whether human behavior, at nation-scale, can be associated with contextual variables such as income, segregation, or education. In this project, we elaborate a geo-computational approach to study whether found correlations between mobile phone and socio-economic indicators are consistent for different definitions of urban areas. Are results suggest that this is not the case, asking for prudence when it comes to interpreting mobile phone indicators in relation to other indicators.
Collaborators on this project are: Clementine Cottineau (CNRS, ex-CASA), Elsa Arcaute (CASA, UCL), and Clement Lee (Open Lab, Newcastle University)
Our paper has been published in the International Journal of Geo-information and become the cover story for their latest issue! What more do you need to go and check it out?
Comparing mobility entropy between regions
In this project, we investigate whether we can use mobile phone data to compare observations on human mobility between large-scale regions in France. We focus on one description of human mobility: mobility entropy, and investigate its spatial distribution in France when calculated for about 18 million French mobile phone users. Our findings show that there is a need to correct mobility entropy measures for cell tower density in order to render comparable results between regions. When done for the French case, we show how the diversity of mobility for French mobile phone users is associated with elements such as land use and car usage.
Collaborators on this project are: Willem Schoors, (KU Leuven), Anton Van Rompeay (KU Leuven), Thomas Ploetz (ex-Open Lab, Georgia Tech), and Zbigniew Smoreda (Orange Labs).
Our paper on the topic can be found here: Comparing Regional Patterns of Individual Movement Using Corrected Mobility Entropy
Sensitivities of home detection practices
Home detection forms a prerequisite step for many analyses of mobile phone data, and other geo-located mobility traces. Nevertheless, validation of such home detection practices is difficult, among others because of privacy laws. As a consequence, no clear information exists on the effect of research choices when opting for one or another home detection algorithm (HDA). In this project, we investigate different HDAs and their performance on a French CDR dataset when compared to data from official statistics. Doing so, we gain insights in the pro's and con's of using different HDAs.
Collaborators on this project are: Benjamin Sakarovitch (INSEE), Clement Lee (Open Lab), and Zbigniew Smoreda (Orange Labs).
Our paper on this topic won the student paper award at the BigSurv'18 conference and will appear in the book Big Data Meets Survey Science in 2020. An arxiv version of this student paper can be found here: Performance and sensitivities of home detection from mobile phone data.
Another paper on this topic was presented at NTTS 2017, and has now been published in the Journal of Official Statistics, if you fancy a read.
Hierarchies in the Belgian logistics system
In this project we investigate how hierarchical structures within the logistics buyer-supplier network in Belgium can be confronted and combined with (qualitative) location- and domain-specific insights. Such research is necessary for understanding, discussing and advancing the role network analysis can play in geography. In our case, we iterate a community detection algorithm on the Belgian logistics network to achieve different levels of communities and to quantify the borders between them. Based on one case study for Antwerp, we show how this type of information can be used to study logistics from a richer geographical perspective.
Collaborators on this project are: Joris Beckers (UAntwerpen) and Ann Verhetsel (UAntwerpen).
Our paper has been published by the Journal of Transport Geography, and therefore lives here: Returning the particular: Understanding hierarchies in the Belgian logistics system
Percolation of interaction and commute networks
In this project, we are exploring how, based on percolation theory, commute and interaction networks derived from mobile phone data can help in understanding the spatial structure of cities in France.
Collaborators on this project are: Carlos Molinero (Complexity Science Hub Vienna) and Elsa Arcaute (CASA, UCL)
We recently took this project to two international conferences: the Conference on Complex Systems (CCS 2019) and the European Colloquiu on Theoretical and Quantitative Geography (ECTQG 2019). Based on the input of our colleagues we aim to finish a first draft of this work by the end of 2019, so stay tuned.
Hierarchical structures of calling patterns
In this project we take a deeper dive into large-scale calling patterns. We analyze, based on mobile phone data, how we can mine country-wide networks of calling for hierarchical structures. Surprisingly, we find that the hierarchical organization of call patterns is similar for different countries in the world. As a consequence, we ask ourselves the question whether this insight can be used to ameliorate existing predictive models, like gravity or radiation models. Personally, it has been an honor to work with this team that brings together several world leading experts in network science, interaction modeling and the use of mobile phone data.
Collaborators are: Sebastian Grauwin (ex-MIT), Michael Szell (MIT), Stanislav Sobolevsky (NYU and MIT), Carlo Ratti (MIT), Philipp Hövel (TUB), Fillipo Simini (Uni. of Bristol), Albert-Laszlo Barabasi (NE Uni.) and Zbigniew Smoreda (Orange Labs).
Find our paper by clicking here: 'Identifying the structural discontinuities of human interactions'
Or read our popular article on the use of mobile phone data to delineate space on Orange's Research Blog
Mobile phone data for Official Statistics
In this joint project between Orange, French Official Statistics Offices (INSEE) and the Statistics Office of the European Commision (Eurostat) we investigate the possibilites, difficulties and methodologies to use mobile phone data for Official Statistics. Working in a team of about 8 people, our work explores, amongst others, the creation of individual indicators, the estimation of population densities and the delineation of urban areas. Our work aims to develop practical insights and knowledge that serve as an important input in decision-making within national statistics offices as well as in European policy governing Big Data.
Collaborators on this project are: Stéphanie Combes (INSEE), Marie-Pierre de Bellefon (INSEE), Benjamin Sakarovitch (INSEE), Pauline Givord (INSEE), Vincent Loonis (INSEE), Fernando Reis (Eurostat), Michail Skaliotis (Eurostat), and Zbigniew Smoreda (Orange Labs).
Find our conference paper here: 'Mining mobile phone data to recognize urban areas',
Or find some of our presentations at the talks page, here: 'Home detection algorithms and mobile phone data', or here: 'Using mobile phone data for official statistics'.
Deriving touristic tours from mobile phone data
This, rather informal, project investigates the possibilities to derive long-distancte trips from mobile phone data. As surveys gathering this kind of information are typically smal-scale and prone to underestimation, large-scale, objective data from mobile phone usage becomes a more than valid alternative. Our obtained results offer new insights in large-scale, long-distance movement that can be used to confront existing statistics, to contribute to models of (long-distance) movement (like done in the PhD-study of Maxim Janzen), to investigate the spatial pattern of domestic tourism trips (like done in the Master thesis of Liane Hendrickx) or to advise policy.
Collaborators on this project are: Maxim Janzen (ETH Zürich), Kay Axhausen (ETH Zürich) Aare Puussaar (Newcastle University, former Positium), Liane Hendrickx (KU Leuven) and Zbigniew Smoreda (Orange Labs).
Find two full papers here: 'Estimating long distance travel demand with mobile phone billing data' and 'Purpose imputation for long-distance tours without personal information'.
A pre-print of a third paper here: 'Closer to the total? Long distance travel of French mobile phone users'
And a presentation here: 'Exploring the use of mobile phones during domestic tourism trips'