Yatao Zhang

Last Name

Zhang

First Name

Yatao

ORCID

0000-0001-5701-2836

Show all metadata

Search Results

Publications 1 - 10 of 15

Counterfactual explanations for deep learning-based traffic forecasting
Wang, Rushan; Xin, Yanan; Zhang, Yatao; et al. (2025)
Communications in Transportation Research
Deep learning models are widely used in traffic forecasting and have achieved state-of-the-art prediction accuracy. However, their black-box nature presents challenges for interpretability and usability, particularly when predictions are significantly influenced by complex urban contextual features. This study aims to leverage an explainable artificial intelligence (AI) approach, counterfactual explanations, to enhance the explainability of deep learning-based traffic forecasting models and elucidate their relationships with various contextual features. We present a comprehensive framework that generates counterfactual explanations for traffic forecasting. The study first implements a graph convolutional network (GCN) to predict traffic speed based on historical traffic data and contextual variables. Counterfactual explanations are generated through a multi-objective optimization process, with four objectives, validity, proximity, sparsity, and plausibility, each emphasizing different aspects of optimization. We investigated the impact of contextual features on traffic speed prediction under varying spatial and temporal conditions. The scenario-driven counterfactual explanations integrate two types of user-defined constraints, directional and weighting constraints, to tailor the search for counterfactual explanations to specific use cases. These tailored explanations benefit machine learning practitioners who aim to understand the model's learning mechanisms and traffic domain experts who seek insights for necessity factors to alter traffic condition. The results showcase the effectiveness of counterfactual explanations in revealing traffic patterns learned by deep learning models and explaining the relationship between traffic prediction and contextual features, demonstrating its potential for interpreting black-box deep learning models.
Context-aware multi-head self-attentional neural network model for next location prediction
Hong, Ye; Zhang, Yatao; Schindler, Konrad; et al. (2023)
Transportation Research Part C: Emerging Technologies
Accurate activity location prediction is a crucial component of many mobility applications and is particularly required to develop personalized, sustainable transportation systems. Despite the widespread adoption of deep learning models, next location prediction models lack a comprehensive discussion and integration of mobility-related spatio-temporal contexts. Here, we utilize a multi-head self-attentional (MHSA) neural network that learns location transition patterns from historical location visits, their visit time and activity duration, as well as their surrounding land use functions, to infer an individual's next location. Specifically, we adopt point-of-interest data and latent Dirichlet allocation for representing locations' land use contexts at multiple spatial scales, generate embedding vectors of the spatio-temporal features, and learn to predict the next location with an MHSA network. Through experiments on two large-scale GNSS tracking datasets, we demonstrate that the proposed model outperforms other state-of-the-art prediction models, and reveal the contribution of various spatio-temporal contexts to the model's performance. Moreover, we find that the model trained on population data achieves higher prediction performance with fewer parameters than individual-level models due to learning from collective movement patterns. We also reveal mobility conducted in the recent past and one week before has the largest influence on the current prediction, showing that learning from a subset of the historical mobility is sufficient to obtain an accurate location prediction result. We believe that the proposed model is vital for context-aware mobility prediction. The gained insights will help to understand location prediction models and promote their implementation for mobility applications.
Intelligent Context-Aware Modeling of Urban Systems and Traffic Dynamics Using Spatiotemporal Data
Zhang, Yatao (2025)
With urban populations surpassing rural ones in 2007 and expected to reach 68% by 2050, global urbanization has climbed to unprecedented heights despite cities covering only about 0.5% of the global land area. This extreme concentration of urban populations has driven up stress on traffic systems, shifting mobility patterns and intensifying congestion. As a result, understanding the complex interactions between urban systems and traffic dynamics proves vital for achieving Sustainable Development Goal 11 and supporting Making Cities Resilient 2030. Meanwhile, the surge in massive spatiotemporal datasets has presented remarkable possibilities for modeling their relationship from the context-aware perspective. This dissertation aims to develop an intelligent context-aware modeling framework that connects urban systems with traffic dynamics through a structured approach involving context sensing, integration, and explanation. The framework can progressively capture, incorporate, and interpret urban contextual information for traffic modeling. It begins by sensing urban systems to investigate their relationship with traffic dynamics, then develops context-aware models that incorporate urban contextual information into traffic prediction, and ultimately explains the underlying mechanisms governing the interaction between urban systems and traffic dynamics. To achieve these objectives, this dissertation introduces a series of computational methods that leverage machine learning techniques to effectively extract and utilize contextual information from spatiotemporal data. Inspired by this context-aware modeling framework, this dissertation makes the following contributions. First, we introduce a geo-semantic framework to construct multi-hierarchical semantic representations of urban contexts and traffic flow, revealing their significant correlations in Singapore. Based on this relationship, we propose two computational methods that integrate contextual information into graph deep learning for traffic speed prediction: one separately learns spatial and temporal contextual representations considering their distinct characteristics, while the other develops a unified context-aware knowledge graph framework for contextual representation learning. Both methods exhibit superior performance compared to baseline models in Singapore. Advancing to the explanation stage, we propose a spatiotemporal causality framework to uncover bidirectional causal influences between urban systems and traffic dynamics across 30 cities, which contributes to achieving more sustainable and resilient cities.
A hierarchical approach for fine-grained urban villages recognition fusing remote and social sensing data
Chen, Dongsheng; Tu, Wei; Cao, Rui; et al. (2022)
International Journal of Applied Earth Observation and Geoinformation
Timely and accurate maps of fine-grained urban villages (UVs) are essential for rational urban planning, which highlights the importance for automatic recognition methods as alternative to labor-intensive land survey, especially for large cities with high-density urban areas where UV maps cannot be updated frequently. However, it is challenging to simultaneously achieve accurate and fine-grained recognition of UVs from remote sensing images in high-density cities, due to the problem of low discrimination of remote sensing features showed in UVs. To address this issue, in this paper, we have proposed a hierarchical recognition framework which can integrate remote and social sensing data to recognize fine-grained UVs. The hierarchical framework follows the human cognition processes and has explicit geographical meaning for each step, which ensures its interpretability. Besides, remote and social sensing data can be fused easily in this framework so that the abstract concept of UV can be sufficiently characterized in both coarse and fine scales. To validate the effectiveness of the proposed approach, extensive experiments in Shenzhen, a typical high-density megacity in China with complicated UVs, have been conducted and a fine-grained map with spatial resolution of 2.5 m was obtained. The results show that the proposed approach achieved an impressive performance, with overall accuracy and Kappa of 96.23% and 0.920 respectively. Furthermore, comparative assessments and ablation studies were performed to demonstrate the effectiveness of the hierarchical recognition framework as well as the fusion of remote and social sensing data.
A global North-South division line for portraying urban development
Zhang, Yatao; Li, Xia; Wang, Shaojian; et al. (2021)
iScience
Rapid urbanization has tremendously changed the global landscape with profound impacts on our society. Nighttime light (NTL) data can provide valuable information about human activities and socioeconomic conditions thus has become an effective proxy to measure urban development. By using NTL-derived urban measures from 1992 to 2018, we analyzed the spatiotemporal patterns of global urban development from country to region to city scales, which presented a distinct North-South divergence characterized by the rising and declining patterns. A global North-South division line was identified to partition the globe into the Line-North and the Line-South geographically, which accorded with the socioeconomic difference from the aspects of urban population and economy. This line may keep a certain degree of stability deriving from the trends of population and economic information but also bears uncertainties in the long term.
Street‐level traffic flow and context sensing analysis through semantic integration of multisource geospatial data
Zhang, Yatao; Raubal, Martin (2022)
Transactions in GIS
Sensing urban spaces from multisource geospatial data is vital to understanding the transportation system in the urban context. However, the complexity of urban context and its indirect interaction with traffic flow deepen the difficulty of exploring their relationship. This study proposes a geo-semantic framework first to generate semantic representations of multi-hierarchical urban context and street-level traffic flow, and then investigate their mutual correlation and predictability using a novel semantic matching method. The results demonstrate that each street is associated with its multi-hierarchical spatial signatures of urban context and street-level temporal signatures of traffic flow. The correlation between urban context and traffic flow displays higher values after semantic matching than those in multi-hierarchies. Moreover, we found that utilizing traffic flow to predict urban context results in better accuracy than the reversed prediction. The results of signature analysis and relationship exploration can contribute to a deeper understanding of context-aware transportation research.
Exploring the spatial differentiation of urbanization on two sides of the Hu Huanyong Line – based on nighttime light data and cellular automata
Chen, Dongsheng; Zhang, Yatao; Yao, Yao; et al. (2019)
Applied Geography
Leveraging context-adjusted nighttime light data for socioeconomic explanations of global urban resilience
Zhang, Yatao; Song, Siqi; Li, Xia; et al. (2024)
Sustainable Cities and Society
Urban resilience, crucial for achieving sustainable development goals, entails the adaptation and recovery of urban systems from external shocks, such as the recent global pandemic. To investigate the response of urban systems to COVID-19, this study employs context-adjusted nighttime light data to model global urban resilience by combining an evolving urban resilience metric (EURm) with the shape similarity of resilience curves. Random effect analysis and counterfactual explanations are then implemented to explore the socioeconomic impacts on urban resilience during the pandemic, providing strategic insights for improvement. Our results delineate five diverse urban resilience patterns, each characterized by distinct phases of downturn, minimum, and recovery, with examples from Malaysia, Japan, the United States, China, and South Sudan. We also find notable correlations between socioeconomic factors and urban resilience, highlighting that stringent measures may reduce resilience, whereas proactive health and containment strategies could bolster resilience. Meanwhile, economic stress reflected by inflation adversely affects resilience. Furthermore, we delve into strategic socioeconomic modifications to enhance urban resilience using counterfactual explanations, underlining the importance of customized interpretation for varying countries. Overall, this study advances our understanding of urban resilience during global crises, guiding context-specific resilience strategies in urban planning and policy-making.
Towards SDG 11: Large-scale geographic and demographic characterisation of informal settlements fusing remote sensing, POI, and open geo-data
Tu, Wei; Chen, Dongsheng; Cao, Rui; et al. (2024)
ISPRS Journal of Photogrammetry and Remote Sensing
Informal settlements’ geographic and demographic mapping is essential for evaluating human-centric sustainable development in cities, thus fostering the road to Sustainable Development Goal 11. However, fine-grained informal settlements’ geographic and demographic information is not well available. To fill the gap, this study proposes an effective framework for both fine-grained geographic and demographic characterisation of informal settlements by integrating openly available remote sensing imagery, points-of-interest (POI), and demographic data. Pixel-level informal settlement is firstly mapped by a hierarchical recognition method with satellite imagery and POI. The patch-scale and city-scale geographic patterns of informal settlements are further analysed with landscape metrics. Spatial-demographic profiles are depicted by linking with the open WorldPop dataset to reveal the demographic pattern. Taking the Guangdong-Hong Kong-Macao Greater Bay Area (GBA) in China as the study area, the experiment demonstrates the effectiveness of informal settlement mapping, with an overall accuracy of 91.82%. The aggregated data and code are released (https://github.com/DongshengChen9/IF4SDG11). The demographic patterns of the informal settlements reveal that Guangzhou and Shenzhen, the two core cities in the GBA, concentrate more on young people living in the informal settlements. While the rapid-developing city Shenzhen shows a more significant trend of gender imbalance in the informal settlements. These findings provide valuable insights into monitoring informal settlements in the urban agglomeration and human-centric urban sustainable development, as well as SDG 11.1.1.
Uncovering nonlinear urban factors of homelessness: Evidence from New York City using interpretable machine learning
Yi, Shengao; Tu, Wei; Zhao, Tianhong; et al. (2025)
Environment and Planning B: Urban Analytics and City Science
Urban homelessness is a complex issue rooted in structural inequalities and spatial disparities, significantly affecting urban life and well-being. Existing research often relies on survey-based or linear regression methods, which are limited in scope, coverage, and their ability to capture nonlinear associations. This study addresses these gaps by combining homeless incident reports from New York City's 311 service with multi-source urban big data and employing a Light Gradient Boosting Machine (LightGBM) model alongside SHapley Additive Explanations (SHAP). Through a census tract-level analysis, we examine how socioeconomic, built environment, transportation, and urban landscape factors relate to homelessness incidence. Our findings show that (1) the importance of predictive factors varies across location types, for instance, information, and communication POIs are most predictive in commercial areas, while felony crime and median income dominate in residential zones; (2) socioeconomic and built environment features are consistently more important than transportation and visual landscape indicators; and (3) many factors exhibit nonlinear relationships and threshold effects, such as sharp increases in homelessness beyond a median rent of $1800 or Gini index of 0.53. These findings offer new insights into the spatial distribution and drivers of homelessness and underscore the value of interpretable machine learning in urban analytics. By identifying key environmental thresholds, this study provides evidence-based guidance for spatially targeted urban interventions, such as prioritizing support services in high-risk areas and designing inclusive public spaces that can help mitigate homelessness and promote more sustainable and equitable cities.

Publications 1 - 10 of 15

Yatao Zhang

Last Name

First Name

ORCID

Organisational unit

Filters

Search Results