On Assessing Usability and Reliability of Anonymized Spatio-Temporal Data

Abstract

Spatio-temporal datasets, particularly mobile network data such as Charging Data Records (CDRs), are widely used to analyze human activities across domains like urban planning, transportation, and infrastructure optimization. However, anonymization and collection biases can compromise data reliability and lead to flawed conclusions if not properly assessed. This paper highlights the importance of a preliminary data characterization phase before reuse. We introduce a generic, reusable methodology for evaluating the usability of anonymized spatio-temporal datasets, structured around three key dimensions: dataset overview, traffic pattern analysis, and mobility pattern analysis. We apply this framework to a large-scale anonymized CDR dataset from Shenzhen, uncovering critical insights into its spatial representativeness, behavioral realism, and application-specific limitations. Our approach supports more informed and context-aware use of spatio-temporal data across diverse domains.