Road user behavior: Describing, inferring, predicting and beyond

Abstract

Data collection and analysis perspectives differ widely across disciplines. In the study of road user behavior, data can be used to explore characteristics of the road user, verify theories, or predict road user behavior. The choice of descriptive, inferential, or predictive analysis greatly depends on the research question(s), which affects the data processing and visualization needs. In this paper, we present differences among the three analysis approaches with examples using a subset of the Strategic Highway Research Program (SHRP2) naturalistic driving data. The findings showcase 11 guidelines that should be considered irrespective of the analysis type.

Keywords

• Statistical modeling
• Machine learning
• Experimental design
• Transportation analysis