Publications

For the full and most up-to-date record, see Google Scholar.

Under review

  • Xia, J., Zhu, Y., Abdelrahman, M., & Biljecki, F. Shaping happiness: How transformations of the visual environment influence urban emotions. Under review.

Journal articles

  1. Liu, H., Yang, S., Abdelrahman, M., Zhu, Y., Wei, X., & Biljecki, F. (2026). Inferring urban functions from Google Maps reviews: A multi-scale, multi-modal and cross-city approach. Computers, Environment and Urban Systems, 102475. DOI

    Summary

    Proposes a framework that uses Google Maps place reviews as a new data stream for inferring urban functions. Text and images from reviews are embedded with BERT and Vision Transformer models, so points of interest in Singapore and Hong Kong can be compared in one shared space. Review volume serves as an indicator of activity intensity, and functional classification is done at three spatial scales (a 1 km hexagonal grid, administrative areas and traffic analysis zones) using graph neural networks with k-means clustering.

    Keywords: Urban functional classification; Multimodal representation learning; Graph neural networks; Volunteered geographic information

  2. Abdelrahman, M., Macatulad, E., Lei, B., Quintana, M., Miller, C., & Biljecki, F. (2025). What is a digital twin anyway? Deriving the definition for the built environment from over 15,000 scientific publications. Building and Environment, 112748. DOI

    Summary

    Uses natural language processing on more than 15,000 full-text articles, compared with a survey of 52 experts, to find what the literature actually means by a digital twin. Definitions differ across manufacturing, building and urban scales, and two groups of digital twins emerge: high-performance real-time and long-term decision support. Simulation, AI/ML, real-time capability and bi-directional data flow turn out to be not yet mature in built-environment twins. The paper derives one definition for building twins and one for city twins, and offers a reproducible method for updating them.

    Keywords: Digital twin; Terminology; NLP; LLMs; Building digital twin; Urban digital twin; City digital twin

  3. Quintana, M., Gu, Y., Liang, X., Hou, Y., Ito, K., Zhu, Y., Abdelrahman, M., & Biljecki, F. (2025). Global urban visual perception varies across demographics and personalities. Nature Cities. DOI

    Summary

    A global survey of 1,000 participants from five countries and 45 nationalities rating street view images, released as the Street Perception Evaluation Considering Socioeconomics dataset. It shows how gender, age, income, education, ethnicity and personality shape perception on six classic indicators (safe, lively, wealthy, beautiful, boring, depressing) and four new ones (live nearby, walk, cycle, green). Machine-learning models trained on existing global datasets overestimate positive indicators and underestimate negative ones relative to the human responses, which argues for local context.

  4. Quintana, M., Liu, F., Torkko, J., Gu, Y., Liang, X., Hou, Y., Ito, K., Zhu, Y., Abdelrahman, M., & Biljecki, F. (2026). It is not always greener on the other side: Greenery perception across demographics and personalities in multiple cities. Landscape and Urban Planning, 105618. DOI

    Summary

    Compares how green a street looks to people with how green it measures, using the Green View Index from street view imagery and a survey of 1,000 people in five countries. Perceived and measured greenery correlate moderately everywhere, but perception tends to overestimate measurement. Demographics and personality matter very little; where a person lives is the main factor, and how greenery is arranged in the scene counts for more than how close it is.

    Keywords: Green view index; Street view imagery; Urban sensing; GeoAI; Urban visual perception; Pedestrian greenery

  5. Ito, K., Zhu, Y., Abdelrahman, M., Quintana, M., Hou, Y., Liang, X., Gu, Y., & Biljecki, F. (2025). ZenSVI: An open-source software for the integrated acquisition, processing and analysis of street view imagery towards scalable urban science. Computers, Environment and Urban Systems, 102283. DOI

    Summary

    Presents ZenSVI, a free and open-source Python package covering the whole street view imagery workflow: downloading from platforms such as Mapillary and KartaView, analysing metadata, running computer vision models, converting images to other projections, depth maps and point clouds, plotting, and exporting results. A Singapore case study demonstrates data quality assessment and clustering. The aim is more transparent, reproducible and scalable research with street view imagery.

    Keywords: Street-level imagery; Python package; Computer vision; FAIR; Reproducibility

  6. Abdelrahman, M., Chong, A., & Miller, C. (2022). Personal thermal comfort models using digital twins: Preference prediction with BIM-extracted spatial-temporal proximity data from Build2Vec. Building and Environment, 207, 108532. DOI

    Summary

    Predicts an individual’s thermal preference with spatial context taken from a BIM. Build2Vec turns objects and relations from the model, together with indoor localization and smartwatch-based ecological momentary assessments, into a graph that feeds a classifier. In a real-world test, accuracy improved by 14% to 28% over baselines that use conventional thermal comfort inputs.

  7. Abdelrahman, M., & Miller, C. (2022). Targeting occupant feedback using digital twins: Adaptive spatial-temporal thermal preference sampling to optimize personal comfort models. Building and Environment, 218, 109090. DOI

    Summary

    Asks when and where occupants should be prompted for feedback to avoid survey fatigue. A scenario-based virtual experiment uses BIM-extracted spatial data and a graph neural network to find regions of similar comfort preference and pick good places to trigger a prompt. On two field datasets it gave 18% to 23% higher sampling quality than zone-based and 4x4 m grid-based sampling.

  8. Abdelrahman, M., Zhan, S., Miller, C., & Chong, A. (2021). Data science for building energy efficiency: A comprehensive text-mining driven review of scientific literature. Energy and Buildings, 242, 110885. DOI

    Summary

    A text-mining review of about 30,000 publications retrieved from the Elsevier API. It maps how data sources, data science techniques and building energy applications relate across the building life cycle. Techniques are applied mostly to operation (for example fault detection and diagnosis) and are under-explored in design and commissioning. The paper points to generative adversarial networks for parametric design and transfer learning for optimal operation.

    Keywords: Reference mining; Natural language processing; Data science; Built environment; Building energy efficiency; Word embeddings

  9. Jayathissa, P., Quintana, M., Abdelrahman, M., & Miller, C. (2020). Humans-as-a-sensor for buildings: Intensive longitudinal indoor comfort models. Buildings, 10(10), 174. DOI

    Summary

    Collects intensive longitudinal comfort feedback with micro ecological momentary assessments on a smartwatch: 30 occupants over two weeks gave 4,378 field surveys. Occupants and spaces were clustered by preference tendencies, and those groups were used as features, together with environmental and wearable sensor data, in multi-class models. The best models reached F1-micro scores of 64% for thermal, 80% for light and 86% for noise preference.

    Keywords: Indoor environmental quality; Thermal comfort models; Personalised comfort model; Machine learning; Ecological momentary assessment; Occupant behaviour

  10. Tartarini, F., Schiavon, S., Quintana, M., Abdelrahman, M., Kim, J., & Miller, C. (2022). Personal comfort models based on wearable and environmental data. Indoor Air, 32(10), e13160. DOI

  11. Tarabieh, K., Nassar, K., Abdelrahman, M., & Mashaly, I. (2019). Statics of space syntax: Analysis of daylighting. Frontiers of Architectural Research, 8(3), 311–318. DOI

    Summary

    Extends space syntax, which analyses spatial configuration, with daylighting and glare measures so that space cognition can be studied together with comfort. A mosque layout serves as the case study, analysed with multi-objective optimization, with implications for glare management and daylight potential in design.

  12. Abdelrahman, M., Moustafa, W. S., & Farag, O. M. (2017). Modelling of Egyptian low-cost-housing natural ventilation: Integration of geometry, orientation and street width optimization. Urban Climate, 21, 318–331. DOI

    Summary

    Uses computational fluid dynamics (standard k-epsilon model) to optimise natural ventilation in three prototypes of Egyptian low-cost housing, varying geometry, orientation and the street height-to-width ratio. Pressure on the windward and leeward walls is computed numerically, airflow rates follow from the Bernoulli equation, and in-situ measurements check the calculations. The results indicate that further modifications to these prototypes are needed.

    Keywords: CFD; Low-cost housing; Natural ventilation; Modelling optimization

  13. Moustafa, W. S., Abdelrahman, M., & Hegazy, I. R. (2018). Building performance assessment of user behaviour as a post occupancy evaluation indicator: Case study on youth housing in Egypt. Building Simulation, 11, 389–403. DOI

    Summary

    A post-occupancy evaluation of an Egyptian youth housing prototype, assessing changes users made on their own such as new openings and sunshades. The effects are checked with CFD, thermal and daylight simulation, validated by a wind tunnel test and in-situ daylight measurements. Some behaviours improved thermal comfort by 60% to 87% and daylight efficiency by 31.8% to 41.4%, making user behaviour a usable feedback tool for future designs.

    Keywords: CFD; Residential prototype; Natural ventilation; User behaviour; POE; Daylighting; Building performance

Conference proceedings

  1. Miller, C., Abdelrahman, M., Chong, A., Biljecki, F., Quintana, M., Frei, M., Romanello, L., & Chew, M. Y. L. (2021). The Internet-of-Buildings (IoB): Digital twin convergence of wearable and IoT data with GIS/BIM. Journal of Physics: Conference Series, 2042(1), 012041. DOI

    Summary

    Introduces the Internet-of-Buildings, the convergence of temporal data from IoT devices and wearables with the spatial context of BIM and GIS. A digital twin case study with 17 participants collecting subjective comfort feedback shows that wearable data can be captured within a BIM data environment.

  2. Sood, T., Quintana, M., Jayathissa, P., Abdelrahman, M., & Miller, C. (2019). The SDE4 learning trail: Crowdsourcing occupant comfort feedback at a net-zero energy building. Journal of Physics: Conference Series, 1343(1), 012141. DOI

    Summary

    Describes the SDE Learning Trail, a mobile app at the net-zero energy building SDE4 at NUS that lets visitors learn about its green features while giving comfort feedback. In three months it gathered 1,163 responses on thermal, visual and aural comfort from 616 participants, 79 of whom gave five or more. Occupants are clustered into comfort personality types as a basis for prediction and recommendation systems.

  3. Quintana, M., Abdelrahman, M., Frei, M., Tartarini, F., & Miller, C. (2021). Longitudinal personal thermal comfort preference data in the wild. Proceedings of the 19th ACM Conference on Embedded Networked Sensor Systems (SenSys), 406–412. DOI

    Summary

    A public dataset from a four-week experiment: 17 participants gave about 1,400 thermal preference responses across 17 indoor and outdoor spaces, using a smartwatch app. Environmental variables were monitored in three buildings, indoor location was tracked with a smartphone app, and participants completed background and personality questionnaires. The dataset is available on Zenodo.

    Keywords: Datasets; Thermal comfort; Smart buildings; Longitudinal experiment

  4. Abdelrahman, M., Chong, A., & Miller, C. (2020). Build2Vec: Building representation in vector space. Proceedings of the Symposium on Simulation for Architecture and Urban Design (SimAUD). DOI

    Summary

    Represents a building in vector space. A BIM in IFC format is converted into a labeled property graph, and node2vec with biased random walks learns embeddings that capture semantic similarity between building components, including spatial and spatio-temporal data. A case study on the net-zero energy building SDE4 at NUS shows promising machine learning applications.

    Keywords: Graph embeddings; node2vec; Feature learning; Representation learning

  5. Abdelrahman, M., Zhan, S., & Chong, A. (2020). A three-tier architecture visual-programming platform for building-lifecycle data management. Proceedings of the Symposium on Simulation for Architecture and Urban Design (SimAUD). DOI

    Summary

    Presents a cloud platform with a three-tier architecture that brings together IoT and building management data, building energy simulation, and data analysis and optimisation libraries. It serves three kinds of users: programmers using visual and textual programming, dashboard viewers, and occupants giving feedback. Separating databases, computation and user interfaces gives flexibility, scalability, reusability and lower latency. The platform was in late alpha with an open-source release planned.

    Keywords: Three-tier architecture; Building energy modelling; Building lifecycle; Visual programming

Preprints

  1. Miller, C., Christensen, R., Leong, J. K., Abdelrahman, M., Tartarini, F., Quintana, M., & Chew, M. Y. L. (2022). Smartwatch-based ecological momentary assessments for occupant wellness and privacy in buildings. arXiv preprint. DOI

    Summary

    Adapts an open-source smartwatch ecological momentary assessment platform, previously used for thermal comfort, to wellness topics: COVID-19 risk perception, privacy and distraction in offices, and triggers of movement behaviour. A proof-of-concept study combined micro-surveys with indoor localization. It found a preference for privacy in certain spaces and different infection-risk perception in naturally versus mechanically ventilated spaces.

    Keywords: Field-based survey; Wearables; Longitudinal data; Wellness; Privacy

Software & invention disclosures

  • Abdelrahman, M. (2021). SpaceBrain: Graph neural network platform for spatial indoor environmental quality prediction. Software Invention Disclosure, NUS. Ref. 2021-140.