Sieg, Louise S; (2025) Tessellating the Space-Time Prism: Regionalisation of In-app Location Data for Privacy Protection and Data Preservation. Doctoral thesis (Ph.D), UCL (University College London).

Preview

Text Sieg_10208399_thesis.pdf Download (15MB) | Preview

Abstract

Datasets containing locational information collected by mobile phones are increasingly used in social science research and mobility analysis, yet they continue to present a range of technical, financial, and ethical challenges. The risk of disclosure of personally identifiable information is a foundational concern in this context, and so to alleviate it, datasets are often aggregated to pre-defined geographic units and presented as counts of the number of mobile devices within them at a given time. The use of grids or units created by statistical agencies for the dissemination of traditional datasets - such as censuses - are common choices for this aggregation process. However, these can result in large variations in the number of devices encapsulated within each geographic unit, resulting in over-generalisation and a loss of information in some areas. Investigating this issue is the core theme of this work, which will explore different regionalisation methodologies and their consequences before alighting on a novel method for generating spatial units tailored for mobile phone data. The central aim is to maximise the granularity of the data, whilst minimising the risks of disclosing personal information. This methodology has applications to widely available datasets and enables bespoke geographical units to be created for different contexts and timescales. The generated units are compared to established aggregates from the England and Wales Census and Ordnance Survey, and assessed through varying temporal granularities. This work seeks to demonstrate that these bespoke outputs minimise data omission caused by low counts and preserve underlying data distribution better than existing aggregation methods. This thesis speaks to the need for data-driven and context-driven regionalisation methodologies in enabling the best use new forms of data in research. It endeavours to contribute to a better understanding and safer use of mobile phone location data in social science and promotes regionalisation as a promising solution to reconcile data granularity with disclosure for sensitive location datasets.

Download activity - last month

Export as JSONXMLCSV

Download activity - last 12 months

Export as JSONCSVXML

Downloads by country - last 12 months

Export as CSVXMLJSON

Archive Staff Only

View Item