As commuting rail networks expand and passenger demand grows, service delays have become a growing challenge, propagating through the network and undermining system reliability. Prevailing research, often reliant on statistical correlations or 'black-box' predictive models, fails to reveal the causal mechanisms of delay propagation. To address this gap, this study proposes a network-centric approach grounded in causal inference to explicitly map the directional pathways of delay.Focusing on the Stockholm commuter rail system, we employ the Peter and Clark Momentary Conditional Independence (PCMCI) algorithm on real-world time-series data to construct a Delay Propagation Causal Network (DPCN). Our multi-stage analysis of the DPCN reveals a highly structured network where delays propagate along stable, predictable pathways. A novel classification identifies four distinct station roles with a clear core-periphery spatial logic. To identify the most critical nodes, we introduce a composite causal delay impact index, which integrates causal strength with real-world delay probabilities and successfully identifies high-impact station clusters that align with peak-hour commuter traffic. A final comparison illustrates the advantages of a causality-based approach over correlation-based methods in distinguishing causal propagation links from spurious associations. This study presents a generalizable, causality-based framework and practical tools for transit authorities, offering a data-driven foundation for proactive network management. It enables operators to identify and mitigate systemic vulnerabilities, thereby enhancing the efficiency, reliability, and resilience of commuter rail systems.

Public transportation networks are highly interconnected, where disruptions like traffic congestion propagate bus delays and impact performance. Identifying delay causes is crucial, yet most studies rely on correlation-based methods rather than causal analysis. Attribution methods like the Shapley value quantify factor contributions but often overlook causal dependencies, leading to potential bias. This study uses a causal discovery model to uncover causal relationships between bus delays and various factors (e.g., operational factors, calendar, and weather). Based on this causal graph, an explainable Artificial Intelligence (AI) method quantifies each factor's contribution to delays, focusing on how these contributions vary at different stops along a route. By integrating scheduled route data and real-time vehicle locations, we analyze factor contributions over time and space, exploring various scenarios along the route. Cross-validation is conducted by comparing the importance ranking of factors with the Seemingly Unrelated Regression Equations (SURE). Results show significant variations in factors contributing to delays along the route. Delays at upstream stops propagate downstream, indicating a cascading effect. Operational factors dominate, accounting for 50%–83% of delays. Notably, delays from the preceding two to three stops have a larger impact than just the immediately preceding one stop, and origin delays strongly affect the first half of the route.

Efficient public transportation is crucial for reducing traffic congestion, cutting carbon emissions, and ensuring fair access to jobs and services. With modern technology, we now have access to large amounts of public transport data, including passenger movements, vehicle trajectories, and other sensor-generated information. The knowledge hidden behind this data has significant potential to enhance transportation planning, operations, and control. However, effectively representing and organizing, as well as extracting useful information from such data to address public transportation issues remains challenging.  

Graphical models have gained significant attention for their strengths in data representation, knowledge interconnection, and complex structure visualization. Notably, knowledge graphs and causal graphs are two distinct types of graphical models and are widely applied in various domains (e.g., social network analysis, drug discovery, and recommendation systems, etc.). Knowledge graphs are good at organizing and connecting massive amounts of data and knowledge, revealing complex relationships, and enabling knowledge mining and inference (answering `what' and `how' questions). Causal graphs are powerful for identifying and analyzing causal relationships, allowing for a deeper understanding of the underlying mechanisms that drive observed data patterns  (answering `why' questions). 

Specifically, the thesis aims to propose two data-driven graphical models (i.e., the knowledge graph and causal graph) and explore their application scenarios in public transportation. It constructs a mobility knowledge graph to represent and organize mobility data, mine travel patterns between stations, and validate its value in trip destination inference and user-station attention estimation. Then, to gain a deeper understanding of transportation operations, the thesis develops causal discovery models for static data to infer causal relationships and generate causal graphs to analyse the variables causing bus delays. Based on the causal graph, it quantifies the contribution of each variable while considering the causal relationships to support the development of target strategies to mitigate delays. Additionally, the thesis also develops a time series causal discovery model to understand bus delay propagation patterns and effects within the public transportation system from a system perspective.

Papers I and II focus on data organization and knowledge inference, construct a mobility knowledge graph (MKG), and explore its applications in public transportation. Paper I introduces the concept of MKG and proposes a framework for constructing it from smart card data by capturing spatiotemporal travel patterns between stations using both rule-based and neural network-based decomposition methods. It validates the MKG framework and demonstrates its value in inferring trip destinations using only tap-in records. Paper II explores another transportation application, proposing a method to estimate the `real' user-station attention from partially observed station visit counts data. It utilizes the MKG to capture latent spatiotemporal travel dependencies between stations to enhance the estimation process by addressing missing values and cold start problems. The framework is validated with both synthetic and real-world data, demonstrating the value of MKG in user-station attention estimation.

Papers IV-VI focus on the research of causal graphs and their applications in public transportation. Before conducting the causal analysis for bus delay, Paper III conducts an empirical study examining the heterogeneous effects of various factors on bus arrival delays. Paper IV focuses on the operational variables and develops causal discovery methods for static data to analyse the variables causing bus delays and evaluate their performance from statistical data fitting and causality interpretation perspectives. It identifies the optimal causal discovery method for analysing the causes of bus delays. Further, based on the causal graph generated in Paper IV, Paper V develops a causality-based Shapley value approach to quantify the contribution of each variable to bus delays to support efficient transportation decision-making. The results are cross-validated with the conventional model (e.g., regression models) to reveal the difference between correlation-based and causality-based analysis approaches. Moreover, Paper VI develops a time series causal discovery model to infer causal relationships between bus stops and generate the spatiotemporal delay propagation causal graph from time series bus stop delay data. Then, it incorporates complex network theory to analyse the bus delay propagation patterns and effects within the public transportation system. 

Effektiv kollektivtrafik är avgörande för att minska trängsel, minska koldioxidutsläppen och säkerställa rättvis tillgång till jobb och tjänster. Med modern teknik har vi nu tillgång till stora mängder kollektivtrafikdata, inklusive passagerarrörelser, fordonsrörelser och sensorgenererad information. Den kunskap som döljs bakom dessa data har stor potential att förbättra transportplanering, drift och styrning. Att effektivt representera och organisera, samt att extrahera användbar information från sådan data för att ta itu med kollektivtrafikproblem är fortfarande en utmaning. 

Grafiska modeller har fått stor uppmärksamhet för sina styrkor inom datarepresentation, kunskapssammankoppling och visualisering av komplexa strukturer. Kunskapsgrafer och kausala grafer är två distinkta typer av grafiska modeller och allmänt tillämpade inom olika domäner (t.ex. sociala nätverksanalyser, läkemedelsutveckling och rekommendationssystem, etc.). Kunskapsgrafer är bra på att organisera och koppla samman enorma mängder data och kunskap, avslöja komplexa samband och möjliggöra kunskapsutvinning och inferens (svara på "vad" och "hur"-frågor). Kausala grafer är kraftfulla för att identifiera och analysera orsakssamband, vilket möjliggör en djupare förståelse av de underliggande mekanismerna som driver observerade datamönster (svara på "varför"-frågor). 

Specifikt syftar avhandlingen till att föreslå två datadrivna grafiska modeller (d.v.s. kunskapsgrafen och kausalgrafen) och utforskar deras tillämpningsscenarier i kollektivtrafiken. Den konstruerar en mobilitetskunskapsgraf för att representera och organisera mobilitetsdata, bryta färdmönster mellan stationer och validera dess värde i slutledning av resemål och uppskattning av användarstations uppmärksamhet. Sedan, för att få en djupare förståelse av transportoperationer, utvecklar avhandlingen kausala upptäcktsmodeller för statisk data för att sluta sig till orsakssamband och generera kausala grafer för att analysera variablerna som orsakar bussförseningar. Baserat på kausalgrafen kvantifierar den bidraget från varje variabel samtidigt som orsakssambanden beaktas för att stödja utvecklingen av målstrategier för att mildra förseningar. Dessutom utvecklar avhandlingen också en tidsseriemodell för orsaksupptäckt för att förstå bussfördröjningsutbredningsmönster och effekter inom kollektivtrafiksystemet ur ett systemperspektiv.

Paper I och II fokuserar på dataorganisation och kunskapsinferens, och konstruerar en mobilitetskunskapsgraf (MKG) och utforskar dess tillämpningar i kollektivtrafik. Artikel I introducerar konceptet MKG och föreslår ett ramverk för att konstruera det från smartkortdata genom att fånga spatiotemporala färdmönster mellan stationer med både regelbaserade och neurala nätverksbaserade nedbrytningsmetoder. Det validerar MKG-ramverket och demonstrerar dess värde i att sluta resmål med hjälp av enbart tap-in-poster. Paper II utforskar en annan transportapplikation, och föreslår en metod för att uppskatta den "riktiga" användarstationens uppmärksamhet från delvis observerade stationsbesöksdata. Den använder MKG för att fånga latenta spatiotemporala resorberoenden mellan stationer för att förbättra uppskattningsprocessen genom att ta itu med saknade värden och kallstartsproblem. Ramverket är validerat med både syntetiska och verkliga data, vilket visar värdet av MKG vid uppskattning av användarstations uppmärksamhet.

Paper IV-VI fokuserar på forskning av kausala grafer och deras tillämpningar i kollektivtrafiken. Innan man genomför orsaksanalysen för bussförseningar, genomför Paper III en empirisk studie som undersöker de heterogena effekterna av olika faktorer på bussens ankomstförseningar operativa variabler och utvecklar kausala upptäcktsmetoder för statiska data för att analysera de variabler som orsakar bussförseningar och utvärdera deras prestanda utifrån statistisk dataanpassning och kausalitetstolkningsmetoden för att analysera orsakerna till bussförseningar kausal graf som genereras i Paper IV, Paper V utvecklar en kausalitetsbaserad Shapley-värdesmetod för att kvantifiera bidraget från varje variabel till bussförseningar för att stödja effektivt transportbeslut. Resultaten korsvalideras med den konventionella modellen (t.ex. regressionsmodeller ) för att avslöja skillnaden mellan korrelationsbaserade och kausalitetsbaserade analysmetoder. Dessutom utvecklar Paper VI en tidsseriekausal upptäcktsmodell för att sluta sig till orsakssamband mellan busshållplatser och generera den spatiotemporala fördröjningsutbredningens kausala grafen från tidsseriens busshållplatsfördröjningsdata. Sedan införlivar den komplex nätverksteori för att analysera bussfördröjningens utbredningsmönster och effekter inom kollektivtrafiksystemet.

Understanding human mobility in urban areas is crucial for transportation planning, operations, and online control. The availability of large-scale and diverse mobility data (e.g., smart card data, GPS data), provides valuable insights into human mobility patterns. However, organizing and analyzing such data pose significant challenges. Knowledge graph (KG), a graph-based knowledge representation method, has been successfully applied in various domains but has limited applications in urban mobility. This paper aims to address this gap by reviewing existing KG studies, introducing the concept of a mobility knowledge graph (MKG), and proposing a general learning framework to construct MKG from smart card data. The MKG represents hidden travel activities between public transport stations, with stations as nodes and their relations as edges. Two decomposition approaches, rule-based and neural network-based models, are developed to extract MKG relations from smart card data, capturing latent spatiotemporal travel dependencies. The case study is conducted using smart card data from a heavily used urban railway system to validate the effectiveness of MKG in predicting individual trip destinations. The results demonstrate the significance of establishing an MKG database, as it assists in a typical problem of predicting individual trip destinations for public transport systems with only tap-in records. Additionally, the MKG framework offers potential for efficient data management and applications such as individual mobility prediction and personalized travel recommendations.

Buses in public transit networks often face operational delays due to dynamic conditions such as traffic congestion, which can propagate through transit routes, affecting overall system performance. Understanding the causes of bus arrival delays is crucial for effective public transport management and control. Moreover, understanding the contribution of each factor to bus delays not only aids in developing targeted strategies to mitigate delays but is also crucial for effective decision-making and planning. Traditional research primarily focuses on correlation-based analysis, lacking the ability to reveal the underlying causal mechanisms. Additionally, no studies have considered the complex causal relationships between factors when quantifying their contributions to outcomes in public transport. This study aims to analyze the factors causing bus arrival delays from a causal perspective, focusing on quantifying the causal contribution of each factor while considering their causal relationships. Quantifying a factor's causal contribution poses challenges due to computational complexity and statistical bias from the limited sample size. Using a causal discovery method, this study generates a causal graph for bus arrival delays and employs the causality-based Shapley value to quantify the contribution of each variable. The study further uses the Double Machine Learning (DML) approach to estimate the causal contributions, which provides a consistent and computationally feasible method. A case study was conducted using Google Transit Feed Specification (GTFS) data, focusing on high-frequency bus routes in Stockholm, Sweden. To validate the model, cross-validation was performed by comparing variable importance rankings with traditional models, including Linear Regression (LR) and Structural Equation Modeling (SEM). The comparison shows that results from the causality-based Shapley value significantly differ from those obtained by traditional methods in terms of importance rankings and influence magnitudes. The findings underscore the significant impact of origin delays on bus punctuality, a factor often underestimated in previous studies. Additionally, it demonstrates that employing a causal discovery model can not only infer causal relationships but also reveal direct and indirect effects, which can provide more intuitive explanations. Finally, although the causal results are mathematically and intuitively sound, it is important to further investigate the real causality impact in practice using lab experiments or A/B tests in real-world settings.

Bus transport, characterized by a complex network of routes and stops, frequently experiences delays that can affect the entire system’s reliability, passenger satisfaction, and operational efficiency. Existing research on bus delay propagation predominantly focuses on the route level. They lack a broader network-level perspective, which is essential for fully understanding the complex interactions and delay propagation. Additionally, previous studies typically rely on correlation-based analysis, which may not adequately uncover the underlying causal mechanisms of bus delay propagation. To understand bus delay propagation in the Public Transport System (PTS), this study employs a causality-based model instead of traditional correlation-based analysis to identify causal relationships between bus stops. We introduce a time-series causal discovery model that integrates temporal and spatial features of stop delays to generate a delay propagation causal graph (DPCG). Then, complex network theory and metrics are used to perform topological analysis on the DPCG and identify key bus stops. The case study is conducted using real-time GTFS data from Stockholm, Sweden. The results indicate that stops with more connections significantly influence delay propagation, and the network displays a distinct community structure with mixed connectivity. Moreover, bus stops exhibit different delay propagation patterns during various time periods. During the morning peak, delays primarily propagate to stops in the inner city due to the commuting surge. In the evening peak, however, delays are more widely distributed across central and suburban areas, reflecting the diversity of after-work travel patterns. The study also reveals that delay propagation extends beyond a single route and affects multiple routes. 

Bus transport, characterized by a complex network of routes and stops, frequently experiences delays that can affect the entire system's reliability, passenger satisfaction, and operational efficiency. Existing research on bus delay propagation predominantly focuses on the route level. They lack a broader network-level perspective, which is essential for fully understanding the complex interactions and delay propagation. Additionally, previous studies typically rely on correlation-based analysis, which may not adequately uncover the underlying causal mechanisms of bus delay propagation. To understand bus delay propagation in the Public Transport System (PTS), this study employs a causality-based model instead of traditional correlation-based analysis to identify causal relationships between bus stops. We introduce a time-series causal discovery model that integrates temporal and spatial features of stop delays to generate a delay propagation causal graph (DPCG). Then, complex network theory and metrics are used to perform topological analysis on the DPCG and identify key bus stops. The case study is conducted using real-time GTFS data from Stockholm, Sweden. The results indicate that stops with more connections significantly influence delay propagation, and the network displays a distinct community structure with mixed connectivity. Moreover, bus stops exhibit different delay propagation patterns during various time periods. During the morning peak, delays primarily propagate to stops in the inner city due to the commuting surge. In the evening peak, however, delays are more widely distributed across central and suburban areas, reflecting the diversity of after-work travel patterns. The study also reveals that delay propagation extends beyond a single route and affects multiple routes.

Bus delays pose significant challenges to urban public transportation systems, impacting operational efficiency and incurring substantial costs. Understanding the causative factors behind bus delays is crucial for developing effective strategies to mitigate them. However, existing research predominantly relies on correlational analysis, which falls short of revealing the underlying causal relationships among factors. This study introduces a novel data-driven causality-based modeling approach, combining data-driven causal discovery and structure equation models (SEM), to investigate the causal relationships among various operational factors influencing bus delays. It can automatically generate a causal delay graph that elucidates the interconnectedness and influence of operational factors on bus delays, providing a deeper understanding of the causal mechanism of bus delays. We explored and evaluated the performance of different causal discovery algorithms in generating causal graph from both aspects of the data fitting and causality discovery performance. The SEM model is used to quantify the direct causal effects among factors in the causal graph. The case study is conducted to validate the causal discovery model performance using Google Transit Feed Specification (GTFS) data from high-frequency bus routes in Stockholm, Sweden. The validation results highlight potential of the data-driven causal discovery models in discovering causality relationships and automating the knowledge discovery process, particularly combining with domain knowledge. The empirical findings show the complexity of factors contributing to bus delays and emphasize the importance of integrating causality into the bus delay factor analysis. For example, a high correlation between origin delay and current arrival delay (coefficient = 0.63) doesn't necessarily indicate causation, and a strong causal link from dwell time and arrival delay also does not reflect a high correlation (coefficient = 0.12). The proof of data-driven causal discovery would facilitate the automated and informed decision-making process to optimize bus services towards better efficiency and reliability.

Accurately and timely predicting pedestrian crossing intentions in real-time is critical for operating intelligent vehicles on roads. Although existing models achieve promising accuracy using complex models and video image data, they are constrained for real-time practical use given the high model complexity, time-consuming data preprocessing, and low-quality image data in the wild. To address these, the paper proposes a Spatial-Temporal Attention Graph Convolution Network model for fast pedestrian crossing intention prediction (PedAST-GCN). It uses a lightweight GCN model as the backbone network with simple but robust graph representations of pedestrian crossing intention modality features, including pedestrian pose, bounding box, and vehicle speeds. The model is validated by comparing it with state-of-the-art models on two large-scale public datasets (JAAD and PIE). The results highlight the better performance of the PedAST-GCN model for pedestrian crossing intention prediction in terms of accuracy and computation times. The ablation analysis confirms the value of the backbone layer and graph design, the designed modality features, the effectiveness of attention mechanisms in capturing long-term dependencies (spatial-temporal attention) and fusing heterogeneous features (modality attention), and the robust performance across various observation lengths and in the presence of noisy data.