Investigating the Role of Big Data in Transportation Safety

Abstract

Big data may offer solutions for many challenges for transportation safety, providing more data faster, with higher spatial and temporal resolution. However, researchers and practitioners identify biases in big data that need to be explored and examined before performing data-driven decision-making. Leveraging semi-structured interviews of big data experts, this study includes a quantified analysis of topic frequency and an evaluation of the reliability of concepts through two independently trained coders. To identify the trends in the unstructured textual contents, the research team developed a text mining pipeline to identify trends, patterns, and biases. The study identifies key terms experts use when describing the role of big data in transportation safety, how the terms relate to the big data experts’ language through network plots, and clustering shows a need to focus on sources, quality, analysis, and implementation of big data. Results show value in maintaining the centrality of transportation experts and the public to determine the proper goals and metrics to evaluate transportation safety. Practitioners and researchers can develop new methods to improve population representation with big data, in addition to addressing difficult transportation safety problems. Working ahead of emerging trends and technologies of big data could support further advancements in transportation safety.

Case Study of Trend Mining in Transportation Research Record Articles

“…The relative weight w j for point C j can be denoted as w j = n j / ( nP ) = f j / P . To gain an overall idea of different variants of CA and their applicability, interested readers can consult the authors’ previous studies ( 55–62 ).…”

Uncovering Deep Structure of Determinants in Large Truck Fatal Crashes

“…CA has become more popular in the field of transportation safety research ( 34–44 ). In the presence of rarely occurring variable categories, TCA produces visuals that are clearer and more easily interpreted than those produced by multiple correspondence analysis (MCA).…”