Twitter Interactions as a Data Source for Transportation Incidents

Twitter is a microblogging platform that contains a large amount of publically accessible user generated content. This content consists of short social interactions between users. These interactions often describe day-to-day events, and can include location information, making them potentially suitable for use in transportation-related analysis. This paper evaluates the use of data from public social interactions on Twitter as a potential complement to traffic incident data. We compare incident records from the California Highway Patrol with Twitter messages related to roadway events over the same time period. Relationships between the two datasets are evaluated by visualizing the density of incidents and tweets that coincide near the same location. Additionally, the content of Twitter messages is weighted by its relevance to traffic incidents. This weighting is then compared to the time and space proximity of the message to an incident record to determine if more vivid Twitter messages may correspond to the presence of incidents. Twitter information is interesting because it is inexpensive, readily accessible, has broad geographic coverage, and provides a uniquely passenger-centric perspective. It is expected that this research will lead to a better understanding of the potential for information from Twitter to add context to other traffic measurements as a supplemental data source.