INTERRING SEMANTIC TRAJECTORIES FROM WI-FI TRACKING DATA TO ANALYSE URBAN PHENOLOGY AND TRAVEL BEHAVIOUR

Abstract

Advancements in info-communication technology over the past decade has enabled the streaming of voluminous flows of data from the ubiquitous network of sensors across the city. In order to use these data for urban analytics and decision support systems, useful information such as an individual’s stops and movements need to be deduced from them. A process commonly known as the semantic enrichment of raw trajectories, there has been an extensive literature aimed at identifying such semantic trajectories. Most of these studies are however catered exclusively for GPS and cellular network data, resulting in a literature gap for other data types such as Wi-Fi. Considering the high spatial granularity, sampling frequency and extensive population coverage afforded by Wi-Fi data, this thesis proposes a model that extracts the stays and journeys from an anonymised individual’s raw Wi-Fi trajectory. It also sheds light on the various data quality issues and employs data cleansing and filtering techniques to maintain its validity and completeness. Evaluation of the model with a ground-truth dataset verified its ability to extract semantic trajectories with high temporal and spatial fidelity whilst maintaining a high precision and moderate recall rate. Finally, the model was employed in a university campus to study the urban phenology (i.e. population dynamics) of a single faculty and travel behaviour for a university campus. Students and staff were found to exhibit stark contrasts in urban phenology and travel behaviour patterns, where the former varied with building function, time of day and the University’s academic calendar, and the latter had some resemblance with the gravity model and Tobler’s first law. These results endorse the capability of the model in gleaning useful insights on the spatiotemporal behaviour of people in a large urban setting, thus setting a foundation for future research.