Building Reproducible Workflows using Transportation Data and COVID-19 Data

Abstract

Human mobility analysis research, or the analysis of individuals' daily activity and travel patterns, is an important topic as the results on people's travel patterns are the basis of hundreds of billions’ investment for the nation's transportation infrastructures. With ubiquitous mobile sensors, human mobility analysis has been increasingly performed with an enormous amount of mobile sensor data such as GPS location data generated from mobile devices. Transportation researchers have developed pre-processing and analysis methods to extract mobility patterns from those big mobile sensor data. These analytical methods are often used by transportation researchers who might not be trained in computer science. Thus, it is time-consuming and non-trivial to configure the computing environment required to execute these methods so as to reproduce published results. This thesis develops the Mobility Analysis Workflow, a containerized software tool, to enhance reproducibility of research in the transportation community. Providing an interactive graphical user interface, the Mobility Analysis Workflow allows users to create and customize mobility analysis pipelines, also known as workflows, in which each graphical module represents a data processing task. Furthermore, we conducted benchmarking experiments to identify the computational bottlenecks of our case study workflow. In addition, we expanded the functionality of the Mobility Analysis Workflow by creating new use cases in modeling the spread of the Coronavirus Disease 2019.