Using Georeferenced Twitter Data to Estimate Pedestrian Traffic in an Urban Road Network

Since existing methods to estimate the pedestrian activity in an urban area are data-intensive, we ask the question whether just georeferenced Twitter data can be a viable proxy for inferring pedestrian activity. Walking is often the mode of the last leg reaching an activity location, from where, presumably, the tweets originate. This study analyses this question in three steps. First, we use correlation analysis to assess whether georeferenced Twitter data can be used as a viable proxy for inferring pedestrian activity. Then we adopt standard regression analysis to estimate pedestrian traffic at existing pedestrian sensor locations using georeferenced tweets alone. Thirdly, exploiting the results above, we estimate the hourly pedestrian traffic counts at every segment of the study area network for every hour of every day of the week. Results show a fair correlation between tweets and pedestrian counts, in contrast to counts of other modes of travelling. Thus, this method contributes a non-data-intensive approach for estimating pedestrian activity. Since Twitter is an omnipresent, publicly available data source, this study transcends the boundaries of geographic transferability and scalability, unlike its more traditional counterparts