MSc thesis on spatiotemporal analysis of human mobility at a music festival using anonymized Wi-Fi traces.
Clustering techniques are used in different domains to uncover patterns that would otherwise remain hidden in complex, unstructured datasets, supporting practical applications such as customer profiling, social behavior analysis, and traffic flow optimization. However, these methods are not typically tailored to the study of human mobility in events held at designated locations, such as music festivals and conventions, where the combination of activities and spatio-temporal dynamics introduces additional layers of complexity to the analysis. This study presents customized preprocessing and trajectory clustering adaptations for interpreting anonymous Wi-Fi traces of event attendees, addressing context-specific challenges such as unidentified sources of signals, uneven sample rates, and noisy sequences, all within an unsupervised setting with no ground truth. This strategy is compared with a network science approach that applies community detection to bipartite graphs built from implicit attendee feedback, discussing the distinct perspectives offered by each method. The clustering methods identified groups of attendees who stayed primarily within the two main audience zones and others with more exploratory behavior. Among the exploratory participants, some followed consistent movement patterns between venue areas, while others exhibited more irregular trajectories. These clusters, which also reflect musical preferences, provided different insights from those of the community detection techniques, which identified smaller groups with a stronger focus on music preference. These findings contribute to the understanding of participant behavior and present a methodological approach adaptable to similar event settings.