Published

Abstract

Accurate modeling of individual travel mode choices is vital for developing sustainable, efficient, and equitable transportation systems. Traditional discrete choice models offer behavioral interpretability but struggle to capture complex and nonlinear travel behavior. Machine learning (ML) techniques provide strong predictive performance yet raise concerns regarding interpretability, transferability, and reproducibility. This study presents a systematic literature review of ML applications in travel mode choice covering publications from January 2017 to March 2025. Following the PRISMA 2020 protocol and strict inclusion and quality assessment procedures, 49 studies were selected for full analysis.A structured coding framework guided data extraction across modeling methods, feature design, optimization strategies, evaluation metrics, and transparency practices. Results indicate that ensemble learners and deep neural networks dominate recent research, improving predictive accuracy but often at the cost of reduced behavioral realism. Feature engineering increasingly integrates spatial, attitudinal, and sensor-based data, while temporal and multimodal dimensions remain underexplored. Most studies emphasize internal validation, with limited attention to behavioral plausibility, causal validity, or external transferability. This paper develops a methodological taxonomy and a unified framework for predictive and behavioral evaluation, highlighting protocols for causal identification and external benchmarking. It also outlines the potential of large language models (LLMs) for data curation, hybrid modeling, and research reproducibility.By bridging statistical learning with economic theory, future research can move toward models that are both technically robust and behaviorally coherent. Leveraging emerging AI tools such as LLMs will further enhance transparency, reproducibility, and alignment with the goals of sustainable and resilient mobility systems.

Hassan, Mahbub, Md Emtiaz Kabir, Syeda Tamzida Akter, Saikat Sarkar Shraban, Khairul Salleh Basaruddin, and Md Ashequl Islam. "Machine Learning in Travel Mode Choice Studies: A Systematic Literature Review of Applications, Methods, and Challenges." Results in Engineering (2025): 108140.

Abstract

In recent years, Machine Learning (ML) has emerged as a transformative technology in the field of Autonomous Vehicles and Autonomous Driving, fundamentally reshaping vehicle perception, decision-making, and control mechanisms. The integration of ML techniques is pivotal to advancing the safety, efficiency, and adaptability of AV systems, driving significant academic and industrial interest. Over the past decade, there has been an unprecedented surge in research output exploring diverse ML methodologies tailored to AV applications. However, despite this exponential growth, there remains a critical gap in systematically understanding the evolution, intellectual landscape, research hotspots, and collaborative dynamics of this rapidly expanding domain, limiting scholars’ and practitioners’ ability to grasp the current state of knowledge and identify emerging trends. To fill this gap, this study presents an extensive bibliometric and conceptual structure analysis, offering a consolidated view of the field’s development from 2011 to 2025. The results uncover exponential growth in scholarly output, largely driven by the proliferation of deep learning, reinforcement learning, and federated learning approaches. The intellectual landscape of ML in AVs is organized around five dominant thematic clusters: perception, planning and decision-making, control systems, safety and robustness, and connected or federated systems. Additionally, an in-depth qualitative review of the most influential publications highlights critical methodological contributions while exposing persistent challenges related to model generalization, safety validation, and interpretability. It concludes with a forward-looking agenda, emphasizing interdisciplinary research, robust validation frameworks, and the integration of emerging technologies such as Digital Twins, Edge AI, Quantum Computing, and Large Language Models (LLMs) to enhance intelligence, scalability, reasoning, adaptability, and human–machine collaboration in future AV systems. 

Hassan, M., Islam, M. K., Amin, M. B., & Narupiti, S. (2025). Mapping the machine learning landscape in autonomous vehicles: A scientometric review of research trends, applications, challenges, and future directions. IEEE Access, 13, 182036–182077. https://doi.org/10.1109/ACCESS.2025.3620637 

Abstract

Understanding household vehicle ownership dynamics in the post-COVID-19 era is critical for designing equitable, resilient, and sustainable transportation policies. This study employs an interpretable machine learning framework to model household vehicle ownership using data from the 2022 National Household Travel Survey (NHTS)—the first nationally representative U.S. dataset collected after the onset of the pandemic. A binary classification task distinguishes between single- and multi-vehicle households, applying an ensemble of algorithms, including Random Forest, XGBoost, Support Vector Machines (SVM), and Naïve Bayes. The Random Forest model achieved the highest predictive accuracy (86.9%). To address the interpretability limitations of conventional machine learning approaches, SHapley Additive exPlanations (SHAP) were applied to extract global feature importance and directionality. Results indicate that the number of drivers, household income, and vehicle age are the most influential predictors of multi-vehicle ownership, while contextual factors such as housing tenure, urbanicity, and household lifecycle stage also exert substantial influence highlighting the spatial and demographic heterogeneity in ownership behavior. Policy implications include the design of equity-sensitive strategies such as targeted mobility subsidies, vehicle scrappage incentives, and rural transit innovations. By integrating explainable artificial intelligence into national-scale transportation modeling, this research bridges the gap between predictive accuracy and interpretability, contributing to adaptive mobility strategies aligned with the United Nations Sustainable Development Goals (SDGs), particularly SDG 11 (Sustainable Cities), SDG 10 (Reduced Inequalities), and SDG 13 (Climate Action).

Hassan, M.; Shraban, S.S.; Ahmed, F.; Amin, M.B.; Nagy, Z. Understanding Post-COVID-19 Household Vehicle Ownership Dynamics Through Explainable Machine Learning. Future Transp. 2025, 5, 136. https://doi.org/10.3390/futuretransp5040136 

Abstract

This paper presents a comprehensive review and bibliometric analysis of Large Language Models (LLMs)intransportation, exploring emerging trends, challenges and future research. Understanding their evolution and impact in transportation research is essential. The study used Scopus as the primary data source, applying Bibliometrix, VOSviewer, and Python for performance analysis and science mapping. This study analyzes 161 peer-reviewed articles and reveals a 25.74% annual growth in scholarly output. IEEE Transactions on Intelligent Transportation Systems and IEEE Transactions on Intelligent Vehicles emerge as the most influential journals by publication volume and impact on LLM research. The findings highlight global disparities in research contributions, with China and the United States dominating by publication volume, followed by Germany and Canada, while developing regions exhibit lower scientific productivity. In addition, the study provides qualitative insights by reviewing recent LLM applications in transportation, examining their key contributions, methodological approaches, inherent limitations, and domain-specific challenges. Key research themes focus on autonomous mobility, traffic optimization, and sustainable transportation networks. Despite significant progress, several challenges remain, including decision-making uncertainties, computational scalability, and high energy consumption. Overcoming these challenges requires greater transparency through causal learning, enhanced reasoning via hybrid AI models, and inclusive frameworks that address algorithmic bias and ensure equitable adoption. 

M. Hassan, M. E. Kabir, M. Jusoh, H. K. An, M. Negnevitsky, and C. Li, "Large Language Models in Transportation: A Comprehensive Bibliometric Analysis of Emerging Trends, Challenges and Future Research," IEEE Access, early access, pp. 1-1, 2025, doi:https://doi.org/10.1109/ACCESS.2025.3589319  

Highlights

• An integrated GTFS–GIS approach identifies structural inefficiencies and spatial inequities in the urban bus networks of Johor Bahru and Penang, Malaysia.

• Larkin Sentral exhibits the highest centrality; Terminal Komtar presents strong nodal connectivity.

• Network metrics indicate Johor Bahru’s polycentric and dispersed configuration, contrasted with Penang’s corridor-focused monocentric structure.

• Accessibility analysis highlights limited first-mile connectivity in peripheral areas and overlapping transit services in central corridors.

• The study recommends GTFS-RT for real-time modeling and proposes future research on adaptive planning aligned with SDG 11.2.

Mahbub Hassan, Hridoy Deb Mahin, Ferdoushi Ahmed, Md. Maruf Hassan, Atikur Rahaman, Masuk Abdullah, Assessing Public Transit Network Efficiency and Accessibility in Johor Bahru and Penang, Malaysia: A Data-Driven Approach,Results in Engineering,Volume 27,2025,106126,ISSN 2590-1230, https://doi.org/10.1016/j.rineng.2025.106126. 

Highlights

• A bibliometric review of 283 studies (2014–2024) explores XR in transportation.

• Key themes: AV validation, pedestrian safety, AR navigation, traffic simulation.

• USA, Germany, and China dominate XR outputs with growing interdisciplinary work.

• Major gaps: ecological validity, demographics, cybersecurity, real-world trials.

• Future XR research must stress scalability, cognition, ethics, and ecosystems.

Hassan, Mahbub, Saikat Sarkar Shraban, Md Ashequl Islam, Khairul Salleh Basaruddin, Muhammad Farzik Ijaz, Nur Saifullah Bin Kamarrudin, and Hiroshi Takemura. "Integration of Extended Reality Technologies in Transportation Systems: A Bibliometric Analysis and Review of Emerging Trends, Challenges, and Future Research." Results in Engineering (2025): 105334.https://doi.org/10.1016/j.rineng.2025.105334 

Highlights

• Provides a comprehensive review of machine learning applications in Intelligent Transportation Systems (ITS).

• Conducts a bibliometric analysis to uncover trends, methodologies, and emerging research areas in ITS.

• Highlights key applications of machine learning in traffic prediction, route optimization, and safety management.

• Identifies future research directions, including the integration of deep learning and AI in ITS.

Hassan, M., Al Nafees, A., Shraban, S.S. et al. Application of machine learning in intelligent transport systems: a comprehensive review and bibliometric analysis. Discov Civ Eng 2, 98 (2025). https://doi.org/10.1007/s44290-025-00256-2 

Highlights

• High-Accuracy Forecasting– Random Forest achieved superior performance with R² = 0.9656, accurately capturing over 96% of ridership variance.

• Multifactor Analysis– Key predictors included day type (weekday/weekend), temperature, and public holidays, emphasizing the role of temporal and exogenous factors.

• Comparative Modeling– Machine learning models (Random Forest, XGBoost) significantly outperformed traditional approaches (Linear Regression, ARIMA) in predictive accuracy.

• Operational Planning Insights– Findings support the use of data-driven models for real-time transit planning, especially to manage peak demand during festivals and adverse weather.

• Future Research Pathways– The study advocates for enhanced accuracy through data integration (e.g., passenger flow, ticketing data) and model refinement across transit systems.

A. Al Nafees, M. Hassan, A. Paul, S. S. Shraban and H. Deb Mahin, "Public Transport Ridership Forecasting Using Machine Learning: A Case Study of Rapid Bus KL in Malaysia," 2024 27th International Conference on Computer and Information Technology (ICCIT), Cox's Bazar, Bangladesh, 2024, pp. 3366-3371, https://doi.org/10.1109/ICCIT64611.2024.11021879   

Highlights

• Gap parameters like critical gaps and follow-up times are essential for optimizing traffic flow and safety across roundabout types.

• The study evaluates single-lane, multi-lane, and turbo roundabouts, highlighting the unique dynamics of each configuration.

• Incorporates computer simulations, empirical observations, and mathematical models to analyze gap parameters effectively.

• Traces the evolution of roundabout designs and emphasizes the importance of region-specific adaptations based on traffic conditions and driver behavior.

Hassan, M., Al Nafees, A., Shraban, S. S., Mahin, H. D., Paul, A., & Ki, A. H. (2024). A comprehensive review on the impact of gap parameters for multiple types of roundabouts in different traffic conditions. International Journal of Integrated Engineering (IJIE), 16(9), 1-20. https://doi.org/10.30880/ijie.2024.16.09.001 

Highlights

• Over 50% of users, primarily from lower and middle socioeconomic classes, find ticket prices reasonable, ensuring accessibility.

• Daily and weekly commuters rely on the metro for business or school, with shorter trip durations being a major benefit.

• Peak-hour congestion and cleanliness issues remain key concerns despite overall positive perceptions.

• The metro system improves traffic flow, reduces travel times, and is favorably viewed for safety, accessibility, and environmental benefits.

Hassan, M., Mehjabin, F., & Al Nafees, A. (2024). A study on the impact of the newly constructed metro rail in Dhaka on travel behavior. In Lecture Notes in Civil Engineering. 13th Asia-Pacific Conference on Transportation and the Environment (APTE), National University of Singapore, Singapore, July 8–10, 2024. https://link.springer.com/book/9789819696413 

Highlights

• Rising Research Trend: Scholarly output in ITS big data has grown rapidly over the past decade, signaling strong academic and industrial engagement.

• Emerging Focus Areas: Key themes include traffic prediction, anomaly detection, safety analytics, and IoT integration.

• Global Contributions: China, the U.S., and Canada lead in publications, reflecting strategic investments in smart mobility.

• Impact-Citation Gap: Rapid publication growth contrasts with slower citation rates—pointing to a need for deeper, high-impact studies.

• Neglected Ethical Issues: Privacy, security, and data ethics remain underexplored, underscoring a critical gap in current research.

Hassan, M., Mahin, H.D., Al Nafees, A. et al. Big data applications in intelligent transport systems: a bibliometric analysis and review. Discov Civ Eng 2, 49 (2025). https://doi.org/10.1007/s44290-025-00205-z 

Highlights

• Developed a robust traffic monitoring system for real-time vehicle detection and classification.

• Utilized YOLOv8 with OpenCV for enhanced precision and computational efficiency.

• Achieved remarkable accuracy in real-world highway traffic conditions.

• Proposed a solution to improve road safety and prevent accidents.

Hassan, M., Mahin, H. D., Jusoh, M., Al Nafees, A., & Paul, A. (2024). Vehicle class detection and counting on a Malaysian road using YOLOv8 and OpenCV. IEEE International Conference on Artificial Intelligence in Engineering and Technology (IICAIET), Sabah, Malaysia, August 26–28, 2024. https://doi.org/10.1109/IICAIET62352.2024.10729957

Highlights

• Effectiveness of Metering Signals: The study demonstrated that metering signals significantly improved traffic flow at unbalanced roundabouts. During the 4:30 PM scenario, the West Lane improved from LOS F to LOS D, and the North Lane from LOS F to LOS B when signals were placed at the South Lane with detectors at 240 meters.

• Optimal Detector Placement: The simulation identified that the optimal detector placement for the 12:00 PM scenario was 240 meters from the stop line, while for the 4:30 PM scenario, the best performance was achieved with detectors at 350 meters from the stop line. These placements helped to minimize queue lengths and vehicle delays.

• Reduction in Emissions and Fuel Consumption: After implementing signal control, emissions (CO, NOx, VOC) and fuel consumption were significantly reduced, particularly on the South and East Lanes. For example, CO emissions on the South Lane dropped from 92.9 g to 53.6 g during the 12:00 PM scenario.

• Statistical Validation: The VISSIM model was statistically validated using RMSE and paired t-tests. The results showed a notable reduction in vehicle delay and emissions, with p-values < 0.05, confirming the significant impact of the metering signals.

Hassan, M., An, H. K., Mahin, H. D., Al Nafees, A., Shraban, S. S., & Paul, A. (2025). Investigating the performance of metering methods in managing unbalanced roundabouts using VISSIM. ASEAN Journal of Scientific and Technological Reports, 28(2), e255674-e255674.