Big data presents unprecedented opportunities to address today’s challenges in mobility and transportation, enabling data-driven insights and innovations. This dissertation explores the application of one big data source, Aggregated Mobile Phone Data (AMPD), as a cost-effective, scalable solution for monitoring and managing urban mobility, using the city of Milan as the empirical setting. The research is divided into two main research questions: one concerning the general applications of AMPD for urban mobility analysis and the other one focusing on the specific use of AMPD for event-based crowd modelling. The first segment highlights the potential uses of AMPD in different contexts: from studying users' distribution over time and space to distinguishing the characteristics of users that populate different areas of the city and the capability of identifying brutal changes and anomalies in users' presence. Diverse analytical methods, including descriptive statistics, clustering, and anomaly detection, were employed to offer valuable insights into user travel behaviour and mobility patterns. The second segment addresses the application of predictive models to assess the impact of large-scale events on urban mobility. Ensemble learning techniques, such as Random Forest and Gradient Boosting, were employed alongside traditional statistical models to capture event-induced variations in urban mobility. The results reveal that AMPD effectively captures spatial-temporal mobility trends across Milan, with clear daily patterns and pronounced peaks in user density during major events. The implemented models accurately identified user surges in areas like San Siro on match days and Fashion Week zones, with Random Forest and Gradient Boosting outperforming traditional models in handling complex, non-linear interactions between time, location, and event factors. These insights underscore AMPD’s potential as a powerful tool for urban mobility management, enabling data-driven decision-making for planners and policymakers.
I big data offrono opportunità senza precedenti per affrontare le sfide odierne nel campo della mobilità e dei trasporti. Questa tesi esplora l'applicazione di una fonte di big data, i dati aggregati provenienti da telefoni cellulari (Aggregated Mobile Phone Data - AMPD), come soluzione per il monitoraggio e la gestione della mobilità urbana, utilizzando la città di Milano come contesto empirico. L'analisi è suddivisa in due principali domande di ricerca: una riguardante le applicazioni generali degli AMPD per lo studio della mobilità urbana e l’altra concentrata sull'uso specifico degli AMPD per modellare le presenze degli utenti in aree dove si svolgono eventi di grande portata. La prima parte della tesi evidenzia i potenziali utilizzi degli AMPD in vari contesti: dallo studio della distribuzione degli utenti alla distinzione delle caratteristiche degli utenti che popolano diverse aree della città, nonché la capacità di individuare cambiamenti bruschi e anomalie. Diversi metodi analitici, tra cui statistiche descrittive, clustering e anomaly detection, sono stati impiegati per offrire supporto nell'applicazione dei vari utilizzi. La seconda parte della tesi affronta l’implementazione di modelli per valutare l'impatto di grandi eventi sulla mobilità urbana. Tecniche di ensemble learning, come Random Forest e Gradient Boosting, sono state utilizzate insieme a modelli statistici tradizionali per catturare le variazioni nella mobilità urbana indotte dagli eventi. I risultati rivelano che gli AMPD catturano efficacemente le tendenze spaziali e temporali della mobilità a Milano, con chiari pattern stagionali e picchi di presenze di utenti durante gli eventi. I modelli implementati sono in grado di identificare accuratamente i flussi di utenti in aree come San Siro nei giorni di avvenimenti sportivi, così come le zone maggiormente frequentate durante la Fashion Week. I risultati ottenuti da questo elaborato sottolineano il potenziale degli AMPD come strumento per la gestione della mobilità urbana, abilitando decisioni informate per pianificatori e responsabili di mobilità e trasporti.
Understanding urban mobility through aggregated mobile phone data: a case study of the City of Milan
Cecconi, Lorenzo
2023/2024
Abstract
Big data presents unprecedented opportunities to address today’s challenges in mobility and transportation, enabling data-driven insights and innovations. This dissertation explores the application of one big data source, Aggregated Mobile Phone Data (AMPD), as a cost-effective, scalable solution for monitoring and managing urban mobility, using the city of Milan as the empirical setting. The research is divided into two main research questions: one concerning the general applications of AMPD for urban mobility analysis and the other one focusing on the specific use of AMPD for event-based crowd modelling. The first segment highlights the potential uses of AMPD in different contexts: from studying users' distribution over time and space to distinguishing the characteristics of users that populate different areas of the city and the capability of identifying brutal changes and anomalies in users' presence. Diverse analytical methods, including descriptive statistics, clustering, and anomaly detection, were employed to offer valuable insights into user travel behaviour and mobility patterns. The second segment addresses the application of predictive models to assess the impact of large-scale events on urban mobility. Ensemble learning techniques, such as Random Forest and Gradient Boosting, were employed alongside traditional statistical models to capture event-induced variations in urban mobility. The results reveal that AMPD effectively captures spatial-temporal mobility trends across Milan, with clear daily patterns and pronounced peaks in user density during major events. The implemented models accurately identified user surges in areas like San Siro on match days and Fashion Week zones, with Random Forest and Gradient Boosting outperforming traditional models in handling complex, non-linear interactions between time, location, and event factors. These insights underscore AMPD’s potential as a powerful tool for urban mobility management, enabling data-driven decision-making for planners and policymakers.File | Dimensione | Formato | |
---|---|---|---|
2024_12_Cecconi_Thesis_01.pdf
accessibile in internet per tutti
Descrizione: Thesis
Dimensione
3.59 MB
Formato
Adobe PDF
|
3.59 MB | Adobe PDF | Visualizza/Apri |
2024_12_Cecconi_ExecutiveSummary_02.pdf
accessibile in internet per tutti
Descrizione: Executive Summary
Dimensione
506 kB
Formato
Adobe PDF
|
506 kB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/231102