A temporal-contextual analysis of urban dynamics using location-based data

Multi-level temporal autoregressive modelling of daily activity satisfaction using GPS-integrated activity diary data

“…Missing locations in a time period less than 10 min are assumed to be equally spaced between the start and end points of that period for simplicity. Based on a spatial threshold of 50 m and a temporal threshold of 10 min (time to conduct a meaningful activity), each participant's daily movement trajectory was characterised as a sequence of consecutive stop and movement episodes, similar to the concept of syntactic trajectories in Grinberger and Shoval (2015). These trajectory episodes were then displayed on the website when a participant was filling in the activity diary at the end of each day to reduce recall bias (Kahneman et al 2004).…”

“…A plethora of methods have been proposed for trajectory semantic enrichment processes, usually depending on application contexts and data availability (Parent et al 2013). For instance, Grinberger and Shoval (2015) detailed a process to construct semantic trajectories by coupling raw GPS traces with residents' digital activity diaries, whereby trajectory segments were characterised by activity information. They also demonstrated the usefulness of such semantic trajectories in producing high-level knowledge on urban dynamics and spatial structure.…”

“…Siła-Nowicka et al (2016) augmented GPS trajectories with urban point of interest (POI) data to identify residents' daily activities and travel modes, and to explore the effects of residential location on daily travel mode choices. Whilst methodological advances revolve around new techniques of annotating and segmenting GPS trajectories with external data (Grinberger and Shoval 2015), innovative spatio-temporal visualisation tools (eg Demšar and Virrantaus 2010), and improved data mining approaches (eg Dodge et al 2012, Yang andGidófalvi 2017), the development of generic statistical models to properly analyse semantic GPS trajectories is rather under-researched. Nonetheless, the importance of statistical models in discovering and drawing inferences on relationships between human activity and contextual factors is beyond doubt.…”

“…Treatment of time and scale has been well-recognised as great challenges in the quantitative analysis of GPS trajectory or movement data (eg Long and Nelson 2013, Kwan and Neutens 2014, Purves et al 2014. From the statistical modelling perspective, trajectory segmentshomogeneous sub-trajectories defined by certain criteria such as activities or travels in residents' daily mobility (eg Grinberger andShoval 2015, Siła-Nowicka et al 2016)are irregularly spaced on the time axis with unequal time intervals, and the number of segments per trajectory usually differ. In contrast, standard panel data and spatial panel data techniques (eg Elhorst 2014, Hsiao 2014 are developed based on data with structured, regularly-spaced and discrete time units, and thus it would be problematic to use them to model sequential or temporal correlations between trajectory segments.…”

“…We utilise a unique GPS-integrated activity diary dataset, in which both GPS trajectories of each participant and detailed information on each activity conducted were recorded for a period of seven days, to provide insight into the nexus between activity participation, geographical context and well-being. Similar to the trajectory enrichment process outlined in Grinberger and Shoval (2015), daily trajectories of participants were annotated and segmented by using additional activity diary data. The resulting trajectory segments represent different activities, and drawing on the locational information of trajectory segments, a range of urban form characteristics of the place or context where activities took place are extracted.…”