学术文献库

Mobility analysis, or understanding and modeling of people's mobility patterns in terms of when, where, and how people move from one place to another, is fundamentally important as such information is the basis for large-scale investment decisions on the nation's multi-modal transportation infrastructure. Recent rise of using passively generated mobile data from mobile devices have raised questions on using such data for capturing the mobility patterns of a population because: 1) there is a great variety of different kinds of mobile data and their respective properties are unknown; and 2) data pre-processing and analysis methods are often not explicitly reported. The high stakes involved with mobility analysis and issues associated with the passively generated mobile data call for mobility analysis (including data, methods and results) to be accessible to all, interoperable across different computing systems, reproducible and reusable by others. In this study, a container system named Mobility Analysis Workflow (MAW) that integrates data, methods and results, is developed. Built upon the containerization technology, MAW allows its users to easily create, configure, modify, execute and share their methods and results in the form of Docker containers. Tools for operationalizing MAW are also developed and made publicly available on GitHub. One use case of MAW is the comparative analysis for the impacts of different pre-processing and mobility analysis methods on inferred mobility patterns. This study finds that different pre-processing and analysis methods do have impacts on the resulting mobility patterns. The creation of MAW and a better understanding of the relationship between data, methods and resulting mobility patterns as facilitated by MAW represent an important first step toward promoting reproducibility and reusability in mobility analysis with passively-generated data.