In recent years, the rapid spread of sensor networks, IoT devices, and connected infrastructures has transformed cities into complex, data-rich ecosystems. This exponential increase in data generation has laid the foundation for the concept of the *smart city* — an interconnected and intelligent environment where physical infrastructures, digital technologies, and citizens continuously interact to improve quality of life, sustainability, and operational efficiency. Smart cities emerged as a response to the challenges of urbanization, resource constraints, and the need for sustainable development. Today, they represent true cyber-physical systems, in which real-world phenomena are sensed, processed, and managed in real time through the integration of technologies such as IoT, edge computing, and 5G. These technologies support applications ranging from traffic optimization and environmental monitoring to energy management and public safety. In this context, Artificial Intelligence (AI) and Machine Learning (ML) play a transformative role, enabling the shift from reactive to predictive city management. By analyzing historical and real-time data, ML models can forecast urban system behaviors and support proactive decision-making. These approaches are applicable in several domains, including predictive traffic management, air pollution forecasting, fault detection in water networks, and efficient resource allocation. The transition from static to predictive urban management is crucial to address the increasing complexity of modern cities. AI models trained on heterogeneous data streams — including sensor data, weather conditions, human mobility traces, and social media inputs — enable not only real-time response but also proactive planning. However, implementing predictive intelligence in urban systems involves significant challenges related to data heterogeneity, quality, and privacy, as well as the need to ensure interoperability among diverse systems and platforms. The aim of this research is to explore how machine learning–based predictive models can be effectively applied across different smart city domains — including transportation, environmental monitoring, water management, and anomaly detection — to tackle everyday urban challenges. The thesis seeks to contribute to the development of smart cities that are not only connected and data-driven but also resilient, anticipatory, and human-centered.
Negli ultimi anni, la rapida diffusione di reti di sensori, dispositivi IoT e infrastrutture connesse ha trasformato le città in ecosistemi complessi e ricchi di dati. Questa crescita esponenziale nella generazione di informazioni ha posto le basi per il concetto di smart city, un ambiente interconnesso e intelligente in cui infrastrutture fisiche, tecnologie digitali e cittadini interagiscono costantemente per migliorare la qualità della vita, la sostenibilità e l’efficienza operativa. Le smart cities sono nate come risposta alle sfide dell’urbanizzazione, dei limiti delle risorse e della necessità di uno sviluppo sostenibile. Oggi rappresentano veri e propri sistemi ciber-fisici, in cui fenomeni reali vengono rilevati, elaborati e gestiti in tempo reale grazie all’integrazione di tecnologie come IoT, edge computing e 5G. Tali tecnologie supportano applicazioni che spaziano dall’ottimizzazione del traffico al monitoraggio ambientale, dalla gestione energetica alla sicurezza pubblica. In questo contesto, l’Intelligenza Artificiale (AI) e il Machine Learning (ML) svolgono un ruolo trasformativo, permettendo di passare da una gestione reattiva a una gestione predittiva delle città. Grazie all’analisi di dati storici e in tempo reale, i modelli di ML consentono di prevedere comportamenti dei sistemi urbani e di intervenire in modo proattivo. Questi approcci trovano applicazione in numerosi ambiti, tra cui la gestione predittiva del traffico, la previsione dell’inquinamento atmosferico, il rilevamento di guasti nelle reti idriche e l’allocazione efficiente delle risorse. Il passaggio da una gestione statica a una gestione predittiva è fondamentale per affrontare la crescente complessità dei sistemi urbani. I modelli di AI, addestrati su flussi di dati eterogenei — come dati da sensori, condizioni meteorologiche, tracciamenti di mobilità e input dai social media — consentono non solo reazioni in tempo reale, ma anche pianificazioni preventive. Tuttavia, la realizzazione di un’intelligenza predittiva urbana presenta sfide significative legate all’eterogeneità, alla qualità e alla privacy dei dati, oltre che alla necessità di garantire interoperabilità tra sistemi e piattaforme differenti. L’obiettivo di questa ricerca è esplorare come modelli predittivi basati su machine learning possano essere applicati efficacemente in diversi domini delle smart cities — tra cui trasporti, monitoraggio ambientale, gestione idrica e rilevamento di anomalie — per affrontare le sfide quotidiane delle aree urbane. La tesi mira a contribuire allo sviluppo di città intelligenti non solo connesse e guidate dai dati, ma anche resilienti, predittive e orientate al benessere dei cittadini.
Machine learning for urban intelligence: Building predictive smart cities [Apprendimento automatico per l’intelligenza urbana: Costruire città intelligenti predittive]
BERLOTTI, MARIAELENA
2025
Abstract
In recent years, the rapid spread of sensor networks, IoT devices, and connected infrastructures has transformed cities into complex, data-rich ecosystems. This exponential increase in data generation has laid the foundation for the concept of the *smart city* — an interconnected and intelligent environment where physical infrastructures, digital technologies, and citizens continuously interact to improve quality of life, sustainability, and operational efficiency. Smart cities emerged as a response to the challenges of urbanization, resource constraints, and the need for sustainable development. Today, they represent true cyber-physical systems, in which real-world phenomena are sensed, processed, and managed in real time through the integration of technologies such as IoT, edge computing, and 5G. These technologies support applications ranging from traffic optimization and environmental monitoring to energy management and public safety. In this context, Artificial Intelligence (AI) and Machine Learning (ML) play a transformative role, enabling the shift from reactive to predictive city management. By analyzing historical and real-time data, ML models can forecast urban system behaviors and support proactive decision-making. These approaches are applicable in several domains, including predictive traffic management, air pollution forecasting, fault detection in water networks, and efficient resource allocation. The transition from static to predictive urban management is crucial to address the increasing complexity of modern cities. AI models trained on heterogeneous data streams — including sensor data, weather conditions, human mobility traces, and social media inputs — enable not only real-time response but also proactive planning. However, implementing predictive intelligence in urban systems involves significant challenges related to data heterogeneity, quality, and privacy, as well as the need to ensure interoperability among diverse systems and platforms. The aim of this research is to explore how machine learning–based predictive models can be effectively applied across different smart city domains — including transportation, environmental monitoring, water management, and anomaly detection — to tackle everyday urban challenges. The thesis seeks to contribute to the development of smart cities that are not only connected and data-driven but also resilient, anticipatory, and human-centered.| File | Dimensione | Formato | |
|---|---|---|---|
|
Berlotti_Mariaelena_XXXVIII_tesi.pdf
accesso aperto
Licenza:
Tutti i diritti riservati
Dimensione
6.81 MB
Formato
Adobe PDF
|
6.81 MB | Adobe PDF | Visualizza/Apri |
I documenti in UNITESI sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/20.500.14242/359618
URN:NBN:IT:UNICT-359618