Jul 2026· Journal of Data Analytics and Artificial Intelligence Applications· 0 citations· 20 references
TL;DR
An integrated prediction-and-visualisation pipeline that transforms complex data distributions into actionable visual analytics, such as interpretable station-to-station demand heatmaps via interactive GIS Folium layers is implemented, providing an operationally robust framework to support smart-city transportation management and build more sustainable urban transit systems.
Abstract
Advances in artificial intelligence have led to the widespread use of complicated nonlinear models in various disciplines, including transportation. These models frequently outperform conventional linear models in terms of prediction accuracy. Machine learning and deep learning models have shown great potential for estimating public transit ridership. Consequently, station-pair passenger flow forecast modelling has become crucial to address emerging needs in expanding megacities, providing backbone support for proactive resource allocation, fleet and infrastructure planning, and emergency crowd management. However, traditional statistical models and baseline machine learning algorithms often face critical bottlenecks in capturing highly non-linear passenger dynamics, complex network topologies, and sudden external operational variations simultaneously. In this study, we proposed a versatile methodology using the extreme gradient boosting tree ensemble (XGBoost) to investigate the relationship between built environment factors in station areas and ridership, focusing on the complex spatial-temporal feature engineering frameworks to capture both deep sequential historical trends and non-linear interactions. This article examines real-world passenger flow data collected between January and December 2025. The dataset comprises over 135,000 dense operational records, enabling an exploratory investigation of passenger commuting patterns, the generation of interactive ridership intensity heatmaps via spatial coordinates, and the implementation of an optimised XGBoost model tailored for rail transit ridership. Evaluating the framework against a 20% hold-out validation matrix yielded exceptional predictive synchronisation, achieving an R score of 0.9083, an MAE of 1,236.48 passengers, and an RMSE of 2,388.85 passengers. The framework's underlying decision hierarchy indicates that structural station demand dynamics (45.24%), regional/demographic dynamics (38.12%), and corridor alignment effects (15.15%) overwhelmingly dictate network variance over minor temporal cycles. To bridge the gap between statistical outputs and operational planning, this research implements an integrated prediction-and-visualisation pipeline that transforms complex data distributions into actionable visual analytics, such as interpretable station-to-station demand heatmaps via interactive GIS Folium layers. Ultimately, this framework effectively addresses the non-linear effects of the built environment and temporal heterogeneity on rail transit ridership, providing an operationally robust framework to support smart-city transportation management and build more sustainable urban transit systems.
Accurate metro ridership prediction is crucial for transit planning, particularly as urban rail networks expand and travel patterns vary across different travel scenarios. Existing models often rely on historical ridership data and may have limited adaptability to temporal and spatial heterogeneity. This study proposes a spatial proximity weighted eXtreme gradient boosting (XGBoost) ensemble to predict daily metro ridership under holidays, weekdays, and weekends. By integrating built environment variables and spatial adjacency, the proposed framework constructs local subdata sets using a nearest-neighbor approach and independently trains submodels. The model reduces dependence on continuous historical ridership sequences and is suitable for limited-data conditions. Using urban rail transit (URT) data from Nanjing, the proposed framework is compared with multiscale geographically weighted regression (MGWR) and global XGBoost. The results show that it achieves higher prediction accuracy in terms of
R
2
, mean absolute error (MAE), and root mean squared error (RMSE). Feature importance analysis confirms the model’s interpretability, revealing temporal differences in the effects of variables. The proposed method effectively captures spatiotemporal heterogeneity in ridership dynamics, providing strong generalizability and practical value for urban transit planning and demand forecasting.
Xinyu Hu, Cong Qi, Yunpeng Zhao et al.· Journal of Transportation En...· 0 citations
Accurate short-term travel time prediction is foundational for sustainable urban mobility and intelligent transportation systems on urban arterial corridors, where travel conditions are shaped by interacting traffic, weather, and public transport factors. This study proposes a multi-source machine learning framework for segment-direction-level prediction in the Denizli city center. Floating car data (FCD), Traffic Control Center (TCC) inductive loop detector measurements, historical weather, and public transport indicators were integrated into a 15 min time-segment structure. The final dataset includes 60 segment-direction targets. Performance was evaluated using Linear Regression, Random Forest, LightGBM, and LSTM under a chronological train-validation-test design. Tree-based ensemble models produced the most stable overall performance, with LightGBM and Random Forest yielding similarly low pooled test errors. Segment-level analyses revealed clear spatial and temporal heterogeneity, showing no single model is universally superior across all links. By providing reliable traffic-state information, the framework enables efficient traffic management and may indirectly reduce delay, fuel use, and emissions; these environmental effects were not quantified. SHAP-based interpretation showed that temporal and traffic-state variables dominate predictions, while weather and public transport provide complementary value.
Muhammed Enes Karaoğlan, Yetis Sazi Murat· Sustainability· 0 citations
Almost all operations of a smart transportation or tourism platform are based on reliable travel demand forecasts, including fleet allocation and dynamic pricing. The information that is now at hand to do the task, trip records, accommodation logs, demographic surveys, are large and heterogeneous, a combination of spatial structure and long-range temporal dependence with sequential behavior and tabular properties that cannot be handled by any single model family. In this paper, an intelligent framework is developed as a hybrid one, where four complementary learners are used to operate in parallel: a spatio-temporal graph neural network (ST-GNN) to represent dependencies between travel zones, a Transformer to represent long-horizon temporal patterns, an LSTM to represent sequential mobility dynamics, and an XGBoost to represent structured demographic and cost characteristics. A stacked meta-learner amalgamates the four branch predictions, and a Q-learning agent transforms predictions into adaptive transport and accommodation suggestions with a reward based on satisfaction-minus-cost. The fused model, tested on the New York City TLC taxi-trip corpus with demographic information, achieves lower RMSE than either ARIMA (0.214) or the best single learner (0.169) and higher R² (0.91), and the policy recommendation policy converges in about 300 episodes. In addition to accuracy, the framework provides a formal fusion architecture, a full algorithmic specification, and a statistical validation protocol, and provides a scalable and interpretable platform to intelligent mobility analytics.
Santosh Kumar Sharma, Satish Chander, Piyush Gupta· International journal of com...· 0 citations
To predict the multi-step urban rail transit passenger flow, we propose a dual-graph spatio-temporal network (DSGN). This network can perform station-level passenger flow prediction and is beneficial for real-time operation of the stations. Predicting multiple future time steps remains challenging due to complex spatial dependencies and nonlinear temporal dynamics. The DSGN designed in this paper adopts a decoupled architecture to separately handle the passenger flow signals and time covariates. It performs parallel convolution on the physical topology graph, applies adaptive convolution on the learned graph, and then fuses through gates. The model also incorporates dilated causal convolution and attention pooling to model temporal dynamics. To validate the model, we conducted tests on the rapid transit dataset of the Massachusetts Bay Transportation Authority (MBTA) (121 stations, 30-minute resolution). The results showed that DSGN achieved a Mean Absolute Error (MAE) of 34.67 and weighted average absolute percentage error (WMAPE) of 18.60%, outperforming the competing spatio-temporal baseline models, and having lower variance across seeds. To further explore the model, we also designed ablation experiments, which showed that dual-graph fusion, time attention pooling, and decoupled feature encoding can all improve the model performance, while replacing the learned node embeddings with demographic priors would reduce accuracy and training stability.
Chaohan Zhong, Jiayu Shao· 2026 IEEE International Conf...· 0 citations
Reliable traffic flow forecasting is a core component of intelligent transportation systems; however, many current approaches are still unable to simultaneously model spatial interdependencies and long-term temporal correlations, particularly in cross-sea corridors that exhibit directional heterogeneity and pronounced temporal variability. This study aims to develop an accurate and stable traffic flow prediction framework for cross-sea corridors. To achieve this, an HGTransformer model was proposed that constructed a hypergraph from traffic nodes based on spatial proximity and correlated flow variations, and used hypergraph convolution to extract spatial node representations. These representations were then fed into a Transformer equipped with multi-head self-attention and positional encoding, enabling the model to capture global temporal dependencies in the evolution of traffic flow. Using hourly traffic flow data from the Huangmaohai cross-sea corridor, the model was tested on 1 to 4 h forecasting horizons and compared with long short-term memory (LSTM), multi-layer perceptron (MLP), random forest (RF), support vector regression (SVR), and Bayesian regression (BR) models. The proposed model achieved the best overall performance, with average mean absolute percentage error (MAPE), mean absolute error (MAE), weighted mean absolute percentage error (WMAPE), and root mean square error (RMSE) of 0.178, 13.375, 0.140, and 20.538, respectively. At the 1 h horizon, these values further decreased to 0.172, 12.678, 0.132, and 19.401, while preserving peak–valley structures more accurately under both short- and longer-horizon forecasting. The main contribution of this study lies in the systematic application and validation of the Huangmaohai Corridor dataset, including a reproducible hypergraph construction strategy tailored specifically for this particular infrastructure.
Fan Jiang, Zhiyong Ma, Pumulo Mukozomba et al.· Applied Sciences· 0 citations
For urban ride-hailing services, accurate shortterm demand forecasting is essential for improving vehicle dispatching efficiency, reducing passenger waiting times, and supporting sustainable and intelligent urban mobility management. Short-term ride-hailing demand is shaped by complex couplings among historical demand, temporal regularity, and weather, which makes accurate prediction challenging. Unlike most existing ride-hailing forecasters that rely on graph-intensive or task-specific architectures, this paper proposes a concise Transformer-based method whose unique contribution to ridehailing demand prediction lies in two aspects: (i) a compact multifactor feature set tailored to short-term ride-hailing fluctuation, integrating historical demand, fine-grained temporal attributes, and weather variables; and (ii) explicit modelling of the nonlinear and non-monotonic effects of extreme weather and rainfall intensity on demand within a single Transformer encoder. Experiments on the City of Chicago Transportation Network Providers (TNP) dataset show that the proposed method achieves an MAE of 14.8, an RMSE of 20.1, and a MAPE of 12.1%, outperforming LSTM, Bi-LSTM, and GRU on all three metrics, with relative reductions of 20.9% in MAE, 19.3% in RMSE, and 23.4% in MAPE versus LSTM. These results demonstrate that a concise Transformer with carefully designed multi-factor features can serve as an effective, reproducible baseline for short-term ridehailing demand forecasting, supporting smarter vehicle allocation, lower idle mileage, and reduced energy consumption and carbon emissions in intelligent transportation systems.
Yuanzhe Liu· 2026 2nd International Confe...· 0 citations