Twitter as a spatio-temporal source for incident management

Abstract

Although Dutch highways are monitored extensively, important details about incidents are sometimes lacking or arrive late in the traffic control center. Tweets sent by traffic participants about their experiences on the road could provide useful information in such cases. A difficult task however is to find relevant tweets in the thousands of tweets that are sent each second by Twitter users worldwide. Very interesting in this matter are tweets that are geographically localized by GPS-coordinates, so-called ‘geotagged’ tweets. It is expected that the spatio-temporal characteristics of geotagged tweets can be used to identify incident-related tweets that are sent on or around highways. This thesis addresses the question how useful Twitter is as a source of spatio-temporal information in the domain of incident management. For a period of 5 months geotagged tweets were harvested from the Twitter API and stored in a geographic database. Zonal regularity analysis was used in an attempt to detect traffic-related events in the area around Amsterdam from the database. It was found that geotagged Twitter data is lacking sufficient quantity and quality in order to be a valuable source of spatio-temporal information for incident management.