Please use this identifier to cite or link to this item:
|Title:||Geographical veracity of indicators from mobile phone data : a study of call detail records data in France|
|Abstract:||The study of mobile phone data opens opportunities in many research domains and for many applications. One point of critique is that, within current analyses, mobile phone users are considered uniform and interchangeable. To counter this social atom problem, good research practice demands an increasing contextualization of research results, for example by confrontation with auxiliary datasets or geographical knowledge. The latter forms the starting point of this thesis. The main argument is that there exists a spatial knowledge gap when it comes to the use of indicators derived from mobile phone data. The presented studies assess the geographical veracity of indicators derived from Call Detail Record (CDR) data and the underlying methods used. Based on a CDR dataset of almost 18.5 million users in France captured during a 154-day period in 2007, they show how mobile phone indicators can be constructed for all individual users using big data technologies. Investigation then is on the performance, sensitivity to user choices, and error estimations of home detection methods, which form a primordial step for the aggregation of users in space. Next, a spatial analysis of the popular Mobility Entropy (ME) indicator is performed, revealing its bias to cell tower density, for which a correction is then proposed. Ultimately, the relations between mobile phone indicators, indicators from other data sources and city definitions in France are explored. The main contribution of the thesis is that it reveals multiple limits of the common practices, results, and interpretations that govern mobile phone data research. The presented studies challenge the veracity of mobile phone indicators in different, predominantly geographical, ways and open up discussion on what should be done to improve trustworthiness.|
|Appears in Collections:||School of Computing Science|
Files in This Item:
|Vanhoof M 2018.pdf||Thesis||34.13 MB||Adobe PDF||View/Open|
|dspacelicence.pdf||Licence||43.82 kB||Adobe PDF||View/Open|
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.