Follow Datanami:
April 2, 2021

Analytics Reveal What Influences Sidewalk Traffic

Sidewalk traffic data can be very telling: it can show how walkable – or not – an area is, it can indicate how popular shops or neighborhoods are, and it can indicate whether the urban planning in an area is functioning as intended. Unfortunately, until recently, it’s been difficult to track at scale. Now, researchers from KAUST and Chalmers University have used automated data collection and analytics to understand the factors that influence sidewalk traffic in a given length of street.

Three and a half years ago, the researchers set up detection devices in Amsterdam, London, and Stockholm comprising some 53 neighborhoods and nearly 700 street segments. (“We chose streets that provided a wide mix of street types and density types from each city,” explained David Bolin, a researcher at KAUST.) For three weeks in October 2017, these devices recorded signals – anonymous, of course – from mobile phones that passed by, filtering out any that came in at over six kilometers per hour (around four miles per hour) to capture just likely pedestrian data.

Using an analysis of variance (ANOVA) model, the researchers then parsed out how the pedestrian flow for a given street varied over each day – and what factors influenced that variation. “In a previous study, we found strong links between the total number of people walking on a given street in one day and certain characteristics of the urban environment,” Bolin said. Two of the main characteristics – the total floorspace of a building and its total ground space at street level – are summed up as “built density,” while the third controlling variable they identified was the centrality of the street in question.

“We took advantage of the power of large-scale data collection to determine if these same variables (density and street type) could explain both the full-day counts in different streets and the variations in flow over the day,” says Bolin. 

With this larger dataset – a boon relative to other, similar studies that had struggled with small, inconsistent datasets – the team substantiated their earlier results and identified another important variable. In addition to built density and street type, “attraction variables” (e.g. local markets, transit stops) emerged as a controlling factor for foot traffic.

The researchers compared their results with four common machine learning methods, finding favorable results but noting that “there is room for improvement in capturing the variability in the data, especially between cities.”

“The results provide insights into the importance of street and density types in designing areas with different qualities,” Bolin said. “Accurate predictions for other cities would require more data from multiple cities in different seasons.”