Passively Collected Location Data Shopping Guide: A Literature Review on Data Quality Evaluation

As passively collected location data (PCLD) have been widely introduced to transportation engineering, researchers and practitioners are experiencing difficulties in evaluating and comparing the data quality of different data sources and attaining the most appropriate data sets for various data needs. Since PCLD are passively collected from anonymized device owners as nonprobability sample data, data users have limited or no access to the data generation and collection process. Therefore, they can only rely on the comprehensive and illustrative description of PCLD data sets to assess their data quality and use cases. This paper aims to provide a shopping guide for potential PCLD data users regarding data categories, quality metrics, data caveats, data applications, etc. First, we present a comprehensive literature review on quality evaluation strategies of general data, transportation data, and practice findings. Next, we propose a set of quantitative quality metrics for PCLD data sets, covering all the potential aspects of interest for different use cases. Furthermore, the paper discusses the qualitative impacts of the upstream data processing and filtering on the applicability of PCLD data sets. By comparing the quality metrics of three sample location-based services (LBS) data sets in the US, we further discuss how different metrics could highlight the advantages or display the disadvantages of each data set, and how they suggest the most appropriate data sets for each application need. Finally, we develop a transportation research-oriented framework for identifying and selecting suitable PCLD data sets for both researchers and practitioners. The framework further sheds light on the standardization of the PCLD community and the commercialization of emerging PCLD data sources.