pith. machine review for the scientific record. sign in

arxiv: 2605.01257 · v1 · submitted 2026-05-02 · 💻 cs.AI

Recognition: unknown

Uncertainty-Aware Trip Purpose Inference from GPS Trajectories via POI Semantic Zones and Pareto Calibration

Bo Yang , Haoxuan Ma , Yifan Liu , Zhiyuan Zhang , Chris Stanford , Morgan Sun , Jiaqi Ma
Authors on Pith no claims yet

Pith reviewed 2026-05-09 14:54 UTC · model grok-4.3

classification 💻 cs.AI
keywords trip purpose inferenceGPS trajectoriesPOI semantic zonesPareto optimizationweakly supervised learningactivity detectionmobility modelingtravel surveys
0
0 comments X

The pith

A weakly supervised framework infers trip purposes from GPS trajectories by matching POI semantic zones to survey distributions through Pareto optimization.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper develops a way to label the reasons for stops in large GPS trajectory datasets without needing any individual-level ground truth annotations. It creates neighborhood POI semantic zones, applies distance-weighted likelihoods to assign activity types, and separates inference rules for mandatory activities like commuting from non-mandatory ones like errands. A multi-phase Pareto optimization then adjusts the assignments so that the overall frequencies of activity types, start times, and durations align closely with statistics from household travel surveys. This approach matters for building scalable travel demand models and policy tools from raw mobility traces that are otherwise hard to interpret at city scale.

Core claim

The framework integrates neighborhood-level POI semantic zones with distance-weighted spatial likelihoods, differentiated inference strategies for mandatory and non-mandatory activities, and a multi-phase Pareto optimization that jointly minimizes distributional divergence from household travel survey statistics and maximizes inference reliability without requiring annotated labels. Evaluated on over 81 million staypoints in Los Angeles, the framework reduces activity type frequency Jensen-Shannon distance by 23 percent, start time JSD by 48 percent, and duration JSD by 12 percent relative to a comparable baseline.

What carries the argument

The multi-phase Pareto optimization that tunes inference parameters to reduce divergence from survey distributions while increasing reliability scores.

If this is right

  • Large-scale GPS datasets can be turned into semantically labeled mobility traces without expensive per-trip annotation efforts.
  • Transportation planning models gain improved estimates of when and how long people engage in different activity types.
  • Policy analyses can draw on uncertainty-aware trip purpose data derived directly from raw trajectories across entire cities.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same calibration technique could transfer to other cities provided local household survey data exists to serve as the reference distribution.
  • Updating POI semantic zones dynamically from fresh location data might further tighten the spatial likelihoods in rapidly changing neighborhoods.

Load-bearing premise

Household travel survey statistics accurately represent the true underlying distributions of activity type, start time, and duration in the GPS-derived staypoints.

What would settle it

Obtain even a modest set of individually labeled GPS staypoints and measure whether the framework's inferred distributions match the labeled data more closely than the baseline method.

Figures

Figures reproduced from arXiv: 2605.01257 by Bo Yang, Chris Stanford, Haoxuan Ma, Jiaqi Ma, Morgan Sun, Yifan Liu, Zhiyuan Zhang.

Figure 1
Figure 1. Figure 1: Illustration of the activity type inference task. (a) Raw staypoints extracted from GPS trajectories, where spatiotemporal stop episodes are detected view at source ↗
Figure 2
Figure 2. Figure 2: Three-stage activity inference pipeline. Stage 1 (a) extracts staypoints from raw trajectory data, (b) constructs semantic zones from POIs, and (c) view at source ↗
Figure 3
Figure 3. Figure 3: Distributional alignment comparison across activity type frequency, start time, and duration distributions. view at source ↗
Figure 4
Figure 4. Figure 4: Prediction stability across confidence levels under Gaussian location view at source ↗
read the original abstract

Large-scale GPS trajectory data offer rich observations of human mobility, yet assigning trip purposes to detected stops remains challenging due to the absence of individual-level ground truth, spatial uncertainty from GPS noise and incomplete points of interest (POIs) coverage, and fundamental behavioral differences across trip purposes. We propose a weakly supervised framework integrating neighborhood-level POI semantic zones with distance-weighted spatial likelihoods, differentiated inference strategies for mandatory and non-mandatory activities, and a multi-phase Pareto optimization that jointly minimizes distributional divergence from household travel survey statistics and maximizes inference reliability without requiring annotated labels. Evaluated on over 81 million staypoints in Los Angeles, the framework reduces activity type frequency Jensen-Shannon distance (JSD) by 23%, start time JSD by 48%, and duration JSD by 12% respectively relative to a comparable baseline. The proposed approach provides a scalable and uncertainty-aware path from raw GPS trajectories to semantically annotated mobility data for travel demand modeling and transportation policy analysis.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The manuscript proposes a weakly supervised framework for inferring trip purposes from GPS trajectories. It combines neighborhood-level POI semantic zones with distance-weighted spatial likelihoods, differentiated inference for mandatory versus non-mandatory activities, and a multi-phase Pareto optimization that jointly minimizes Jensen-Shannon divergence (JSD) from household travel survey statistics while maximizing an internal reliability score. Evaluated on more than 81 million staypoints in Los Angeles, the method reports JSD reductions of 23% for activity-type frequency, 48% for start-time distributions, and 12% for duration distributions relative to a comparable baseline.

Significance. If the reported gains can be shown to reflect improved inference rather than calibration artifacts, the framework would offer a scalable, label-free route to semantically annotated mobility data at city scale, directly supporting travel-demand modeling and policy analysis. The evaluation scale is substantial, but the absence of independent ground-truth checks limits the strength of the conclusions.

major comments (2)
  1. [Abstract and §3] Abstract and §3 (Pareto optimization description): the central evaluation metrics (activity-type, start-time, and duration JSD) are identical to the distributional divergences explicitly minimized during the multi-phase Pareto calibration. Because the optimization directly targets the same household-survey marginals used as the evaluation target, the reported 23/48/12 % reductions are consistent with stronger proxy matching rather than with superior trip-purpose inference on the GPS staypoints. An independent check (e.g., sensitivity to held-out survey subsets, alternative aggregate proxies, or cross-city transfer) is required to substantiate the claim that the POI semantic zones and mandatory/non-mandatory logic drive the gains.
  2. [§4] §4 (experimental setup): the baseline is described only as “comparable,” without an explicit enumeration of which components (POI zones, differentiated logic, or Pareto calibration) it includes or excludes. Without this decomposition and without reported error bars, statistical tests, or ablation tables, it is impossible to isolate the contribution of each proposed element to the observed JSD improvements.
minor comments (2)
  1. [§3] Clarify the precise definition and weighting of the “reliability score” maximized in the Pareto phase, including any free parameters and their selection procedure.
  2. [Abstract] The abstract states “over 81 million staypoints” but omits the temporal span, data provider, and filtering criteria; these details are needed for reproducibility.

Simulated Author's Rebuttal

2 responses · 1 unresolved

We thank the referee for the constructive and detailed comments. We address each major comment point by point below, indicating planned revisions where appropriate.

read point-by-point responses

  1. Referee: [Abstract and §3] Abstract and §3 (Pareto optimization description): the central evaluation metrics (activity-type, start-time, and duration JSD) are identical to the distributional divergences explicitly minimized during the multi-phase Pareto calibration. Because the optimization directly targets the same household-survey marginals used as the evaluation target, the reported 23/48/12 % reductions are consistent with stronger proxy matching rather than with superior trip-purpose inference on the GPS staypoints. An independent check (e.g., sensitivity to held-out survey subsets, alternative aggregate proxies, or cross-city transfer) is required to substantiate the claim that the POI semantic zones and mandatory/non-mandatory logic drive the gains.

    Authors: We acknowledge that the evaluation JSD metrics match the optimization targets, raising a legitimate question about whether gains reflect calibration rather than improved inference. The multi-phase Pareto procedure is explicitly multi-objective: it minimizes the three JSD terms while simultaneously maximizing an internal reliability score computed solely from POI semantic zones, distance-weighted likelihoods, and the mandatory/non-mandatory differentiation logic, without using survey marginals. This internal objective is intended to guard against pure proxy matching. To provide the requested independent checks, the revision will add a sensitivity analysis that holds out random subsets of the survey data during calibration and evaluates JSD on the held-out portions. We will also report the achieved reliability scores alongside the JSD values to document the trade-off surface. Cross-city transfer is not feasible with the current Los Angeles-only dataset. revision: partial

  2. Referee: [§4] §4 (experimental setup): the baseline is described only as “comparable,” without an explicit enumeration of which components (POI zones, differentiated logic, or Pareto calibration) it includes or excludes. Without this decomposition and without reported error bars, statistical tests, or ablation tables, it is impossible to isolate the contribution of each proposed element to the observed JSD improvements.

    Authors: We agree that the baseline description and experimental reporting are insufficient to isolate component contributions. In the revised manuscript we will explicitly define the baseline as the full POI-zone plus differentiated-inference pipeline without the Pareto calibration stage. We will add a full ablation table that successively removes POI semantic zones, the mandatory/non-mandatory differentiation, and the Pareto phase, reporting the resulting JSD values. In addition, we will rerun all experiments with multiple random seeds, report standard deviations, and include paired statistical tests on the JSD differences. revision: yes

standing simulated objections not resolved
  • Cross-city transfer validation, which would require GPS trajectories and matching household survey data from a second city not available in the current study.

Circularity Check

1 steps flagged

JSD reductions measured against the same household survey distributions explicitly minimized as calibration target in Pareto phase

specific steps
  1. fitted input called prediction [Abstract (Pareto optimization and evaluation)]
    "a multi-phase Pareto optimization that jointly minimizes distributional divergence from household travel survey statistics and maximizes inference reliability without requiring annotated labels. Evaluated on over 81 million staypoints in Los Angeles, the framework reduces activity type frequency Jensen-Shannon distance (JSD) by 23%, start time JSD by 48%, and duration JSD by 12% respectively relative to a comparable baseline."

    The optimization step explicitly minimizes the identical JSD quantities (activity type frequency, start time, duration) to the household survey statistics that are subsequently used as the ground-truth reference for reporting JSD reductions. Because the evaluation target is the same marginal distribution the calibration phase was constructed to approach, any measured improvement over the baseline is at least partly a direct consequence of how successfully the Pareto phase fitted those survey marginals rather than an external validation of trip-purpose accuracy.

full rationale

The framework's multi-phase Pareto optimization is defined to jointly minimize the exact distributional divergences (activity-type, start-time, and duration JSDs) from household travel survey statistics that later serve as the evaluation benchmark. Reported percentage reductions are therefore computed relative to a baseline on the same target distributions the method was optimized to match. This configuration makes the headline performance numbers consistent with improved proxy fitting rather than an independent test of inference accuracy on the GPS staypoints themselves. The POI semantic zones and mandatory/non-mandatory logic supply additional structure, but the load-bearing claim of superiority rests on the calibrated JSD metrics.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

Abstract-only review; ledger populated from explicit statements in the abstract. The framework rests on the assumption that survey aggregates are a valid calibration target and on several tunable weights inside the Pareto optimizer.

free parameters (1)
  • Pareto phase weights and reliability thresholds
    Multi-phase optimizer balances distributional divergence against inference reliability; specific values are chosen to produce the reported JSD numbers.
axioms (1)
  • domain assumption Household travel survey statistics represent the true marginal distributions of activity type, start time, and duration in the target GPS population.
    Invoked to justify using survey aggregates as both calibration target and evaluation benchmark.

pith-pipeline@v0.9.0 · 5483 in / 1407 out tokens · 58966 ms · 2026-05-09T14:54:57.907543+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

44 extracted references · 2 canonical work pages

  1. [1]

    Human mobility: Models and applications,

    H. Barbosa, M. Barthelemy, G. Ghoshal, C. R. James, M. Lenormand, T. Louail, R. Menezes, J. J. Ramasco, F. Simini, and M. Tomasini, “Human mobility: Models and applications,”Physics Reports, vol. 734, pp. 1–74, 2018

    2018

  2. [2]

    Dynamic traffic flow prediction based on gps data,

    E. Necula, “Dynamic traffic flow prediction based on gps data,” in2014 IEEE 26th international conference on tools with artificial intelligence. IEEE, 2014, pp. 922–929

    2014

  3. [3]

    Mapping global variation in human mobility,

    M. U. Kraemer, A. Sadilek, Q. Zhang, N. A. Marchal, G. Tuli, E. L. Cohn, Y . Hswen, T. A. Perkins, D. L. Smith, R. C. Reiner Jr et al., “Mapping global variation in human mobility,”Nature Human Behaviour, 2020

    2020

  4. [4]

    Learning universal human mobility patterns with a foundation model for cross-domain data fusion,

    H. Ma, X. Liao, Y . Liu, Q. Jiang, C. Stanford, S. Cao, and J. Ma, “Learning universal human mobility patterns with a foundation model for cross-domain data fusion,”Transportation Research Part C: Emerging Technologies, vol. 180, p. 105311, 2025

    2025

  5. [5]

    Trajectory data mining: an overview,

    Y . Zheng, “Trajectory data mining: an overview,”ACM Transactions on Intelligent Systems and Technology (TIST), vol. 6, no. 3, 2015

    2015

  6. [6]

    Editorial special section on coordination, cooperation, and control of autonomous vehicles in smart connected road environments,

    A. Petrillo and S. Santini, “Editorial special section on coordination, cooperation, and control of autonomous vehicles in smart connected road environments,”IEEE Open Journal of Intelligent Transportation Systems, vol. 5, pp. 202–204, 2024

    2024

  7. [7]

    Next-generation travel demand modeling with a generative framework for household activity coordination,

    X. Liao, H. Ma, Y . Liu, Y . Wei, B. Y . He, C. Stanford, and J. Ma, “Next-generation travel demand modeling with a generative framework for household activity coordination,”arXiv preprint arXiv:2507.08871, 2025

    work page arXiv 2025

  8. [8]

    Semantic trajectory data mining with llm-informed poi classification,

    Y . Liu, C. Kuai, X. Liao, H. Ma, B. Y . He, and J. Ma, “Semantic trajectory data mining with llm-informed poi classification,” in2024 IEEE 27th International Conference on Intelligent Transportation Systems (ITSC), 2024, pp. 207–213

    2024

  9. [9]

    Semantic trajectories modeling and analysis,

    C. Parent, S. Spaccapietra, C. Renso, G. Andrienko, N. Andrienko, V . Bogorny, M. L. Damiani, A. Gkoulalas-Divanis, J. Macedo, N. Pelekiset al., “Semantic trajectories modeling and analysis,”ACM Computing Surveys (CSUR), vol. 45, no. 4, pp. 1–32, 2013

    2013

  10. [10]

    Deriving and validating trip purposes and travel modes for multi-day gps-based travel surveys: A large-scale application in the netherlands,

    W. Bohte and K. Maat, “Deriving and validating trip purposes and travel modes for multi-day gps-based travel surveys: A large-scale application in the netherlands,”Transportation Research Part C: Emerging Technologies, vol. 17, no. 3, pp. 285–297, 2009

    2009

  11. [11]

    Deriving rules for trip purpose identification from gps travel survey data and land use data: A machine learning approach,

    Z. Deng and M. Ji, “Deriving rules for trip purpose identification from gps travel survey data and land use data: A machine learning approach,” inTraffic and Transportation Studies 2010, 2010

    2010

  12. [12]

    Assessing the accuracy of the sydney household travel survey with gps,

    P. Stopher, C. FitzGerald, and M. Xu, “Assessing the accuracy of the sydney household travel survey with gps,”Transportation, 2007

    2007

  13. [13]

    An analysis of the factors influencing differences in survey-reported and gps-recorded trips,

    S. G. Bricka, S. Sen, R. Paleti, and C. R. Bhat, “An analysis of the factors influencing differences in survey-reported and gps-recorded trips,”Transportation research part C: emerging technologies, 2012

    2012

  14. [14]

    Identi- fication of activity stop locations in gps trajectories by density-based clustering method combined with support vector machines,

    L. Gong, H. Sato, T. Yamamoto, T. Miwa, and T. Morikawa, “Identi- fication of activity stop locations in gps trajectories by density-based clustering method combined with support vector machines,”Journal of Modern Transportation, vol. 23, no. 3, pp. 202–213, 2015

    2015

  15. [15]

    Transport mode detection based on mobile phone network data: A systematic review,

    H. Huang, Y . Cheng, and R. Weibel, “Transport mode detection based on mobile phone network data: A systematic review,”Transportation Research Part C: Emerging Technologies, vol. 101, pp. 297–312, 2019

    2019

  16. [16]

    Planet dump retrieved from https://planet.osm.org,

    OpenStreetMap contributors, “Planet dump retrieved from https://planet.osm.org,” https://www.openstreetmap.org, 2017

    2017

  17. [17]

    Points of interest (poi): A commentary on the state of the art, challenges, and prospects for the future,

    A. Psyllidis, S. Gao, Y . Hu, E.-K. Kim, G. McKenzie, R. Purves, M. Yuan, and C. Andris, “Points of interest (poi): A commentary on the state of the art, challenges, and prospects for the future,” Computational urban science, vol. 2, no. 1, p. 20, 2022

    2022

  18. [18]

    Habitual travel behaviour: evidence from a six-week travel diary,

    R. Schlich and K. W. Axhausen, “Habitual travel behaviour: evidence from a six-week travel diary,”Transportation, vol. 30, no. 1, 2003

    2003

  19. [19]

    Transportation secure data center,

    National Renewable Energy Laboratory, “Transportation secure data center,” https://www.nrel.gov/tsdc, 2019, accessed: Jan. 15, 2019

    2019

  20. [20]

    Discovering latent activity patterns from transit smart card data: A spatiotemporal topic model,

    Z. Zhao, H. N. Koutsopoulos, and J. Zhao, “Discovering latent activity patterns from transit smart card data: A spatiotemporal topic model,” Transportation Research Part C: Emerging Technologies, 2020

    2020

  21. [21]

    De- ducing mode and purpose from gps data,

    P. Stopher, E. CLIFFORD, J. ZHANG, and C. FITZGERALD, “De- ducing mode and purpose from gps data,”Institute of Transport and Logistics Studies Working Paper, no. ITLS-WP-08-06, 2008

    2008

  22. [22]

    Exploring individual activity-travel patterns based on geolocation data from mobile phones,

    B. Yin and F. Leurent, “Exploring individual activity-travel patterns based on geolocation data from mobile phones,”Transportation Re- search Record, vol. 2675, no. 12, pp. 771–783, 2021

    2021

  23. [23]

    Semantic enhancement of human urban activity chain construction using mobile phone signaling data,

    S. Liu, Y . Long, L. Zhang, and H. Liu, “Semantic enhancement of human urban activity chain construction using mobile phone signaling data,”ISPRS International Journal of Geo-Information, p. 545, 2021

    2021

  24. [24]

    Estimating activity patterns using spatio- temporal data of cell phone networks,

    S. Zahedi and Y . Shafahi, “Estimating activity patterns using spatio- temporal data of cell phone networks,”International Journal of Urban Sciences, vol. 22, no. 2, pp. 162–179, 2018

    2018

  25. [25]

    A combined approach of fuzzy cognitive maps and fuzzy rule-based inference supporting freeway traffic control strategies,

    M. Amini, M. F. Hatwagner, and L. T. Koczy, “A combined approach of fuzzy cognitive maps and fuzzy rule-based inference supporting freeway traffic control strategies,”Mathematics, vol. 10, no. 21, 2022

    2022

  26. [26]

    A random forest model for travel mode identification based on mobile phone signaling data,

    Z. Lu, Z. Long, J. Xia, and C. An, “A random forest model for travel mode identification based on mobile phone signaling data,” Sustainability, vol. 11, no. 21, p. 5950, 2019

    2019

  27. [27]

    Inferring hybrid transportation modes from sparse gps data using a moving window svm classification,

    A. Bolbol, T. Cheng, I. Tsapakis, and J. Haworth, “Inferring hybrid transportation modes from sparse gps data using a moving window svm classification,”Computers, Environment and Urban Systems, vol. 36, no. 6, pp. 526–537, 2012

    2012

  28. [28]

    Towards semantic travel behavior prediction for private car users,

    H. Chen, D. Wang, and C. Liu, “Towards semantic travel behavior prediction for private car users,” in2020 IEEE 22nd International Con- ference on High Performance Computing and Communications; IEEE 18th International Conference on Smart City; IEEE 6th International Conference on Data Science and Systems (HPCC/SmartCity/DSS). IEEE, 2020, pp. 950–957

    2020

  29. [29]

    Graph-aware chained trip purpose infer- ence,

    S. Lyu and T. Kusakabe, “Graph-aware chained trip purpose infer- ence,” in2021 IEEE International Intelligent Transportation Systems Conference (ITSC). IEEE, 2021, pp. 3691–3697

    2021

  30. [30]

    Exploring transformer and graph convolutional networks for human mobility modeling,

    R. Corrias, M. Gjoreski, and M. Langheinrich, “Exploring transformer and graph convolutional networks for human mobility modeling,” Sensors, vol. 23, no. 10, p. 4803, 2023

    2023

  31. [31]

    Sgcn: Sparse graph convolution network for pedestrian trajectory prediction,

    L. Shi, L. Wang, C. Long, S. Zhou, M. Zhou, Z. Niu, and G. Hua, “Sgcn: Sparse graph convolution network for pedestrian trajectory prediction,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2021, pp. 8994–9003

    2021

  32. [32]

    Global and local in- terattribute relationships-based graph convolutional network for flight trajectory prediction,

    Y . Fan, Y . Tan, L. Wu, H. Ye, and Z. Lyu, “Global and local in- terattribute relationships-based graph convolutional network for flight trajectory prediction,”IEEE Transactions on Aerospace and Electronic Systems, vol. 60, no. 3, pp. 2642–2657, 2024

    2024

  33. [33]

    A generative model of urban activities from cellular data,

    M. Yin, M. Sheehan, S. Feygin, J.-F. Paiement, and A. Pozdnoukhov, “A generative model of urban activities from cellular data,”IEEE Transactions on Intelligent Transportation Systems, vol. 19, no. 6, pp. 1682–1696, 2017

    2017

  34. [34]

    Large language models as urban res- idents: An llm agent framework for personal mobility generation,

    J. Wang, R. Jiang, C. Yang, Z. Wu, M. Onizuka, R. Shibasaki, N. Koshizuka, and C. Xiao, “Large language models as urban res- idents: An llm agent framework for personal mobility generation,” Advances in Neural Information Processing Systems, vol. 37, 2024

    2024

  35. [35]

    Agentic large lan- guage models for generating large-scale urban daily activity patterns,

    Y . Zhang, K. Zhang, Y . Pang, and Y . Sekimoto, “Agentic large lan- guage models for generating large-scale urban daily activity patterns,” in2024 IEEE International Conference on Big Data (BigData). IEEE, 2024, pp. 6815–6822

    2024

  36. [36]

    Har- nessing large language models for group poi recommendations,

    J. Long, L. Qu, J. Yu, T. Chen, Q. V . H. Nguyen, and H. Yin, “Har- nessing large language models for group poi recommendations,” in Proceedings of the 34th ACM International Conference on Information and Knowledge Management, 2025, pp. 1968–1978

    2025

  37. [37]

    From gps to semantic data: how and why—a framework for enriching smartphone trajectories,

    A. Ibrahim, H. Zhang, S. Clinch, and S. Harper, “From gps to semantic data: how and why—a framework for enriching smartphone trajectories,”Computing, vol. 103, no. 12, pp. 2763–2787, 2021

    2021

  38. [38]

    Segmenting human trajectory data by movement states while address- ing signal loss and signal noise,

    S. Hwang, C. VanDeMark, N. Dhatt, S. V . Yalla, and R. T. Crews, “Segmenting human trajectory data by movement states while address- ing signal loss and signal noise,”International Journal of Geograph- ical Information Science, vol. 32, no. 7, pp. 1391–1412, 2018

    2018

  39. [39]

    Difftraj: Generating gps trajectory with diffusion probabilistic model,

    Y . Zhu, Y . Ye, S. Zhang, X. Zhao, and J. Yu, “Difftraj: Generating gps trajectory with diffusion probabilistic model,”Advances in Neural Information Processing Systems, vol. 36, pp. 65 168–65 188, 2023

    2023

  40. [40]

    A survey on deep learning for human mobility,

    M. Luca, G. Barlacchi, B. Lepri, and L. Pappalardo, “A survey on deep learning for human mobility,”ACM Computing Surveys (CSUR), vol. 55, no. 1, pp. 1–44, 2021

    2021

  41. [41]

    A density-based algorithm for discovering clusters in large spatial databases with noise,

    M. Ester, H.-P. Kriegel, J. Sander, X. Xuet al., “A density-based algorithm for discovering clusters in large spatial databases with noise,” inkdd, vol. 96, no. 34, 1996, pp. 226–231

    1996

  42. [42]

    A fast and elitist multiobjective genetic algorithm: Nsga-ii,

    K. Deb, A. Pratap, S. Agarwal, and T. Meyarivan, “A fast and elitist multiobjective genetic algorithm: Nsga-ii,”IEEE transactions on evolutionary computation, vol. 6, no. 2, pp. 182–197, 2002

    2002

  43. [43]

    Global mobility and location data provider,

    Veraset, “Global mobility and location data provider,” https://www.ve raset.com/, 2024, accessed: Jun. 17, 2024

    2024

  44. [44]

    Beyond 9-to-5: A generative model for augmenting mobility data of underrepresented shift workers,

    H. Ma, X. Liao, Y . Liu, C. Stanford, and J. Ma, “Beyond 9-to-5: A generative model for augmenting mobility data of underrepresented shift workers,”arXiv preprint arXiv:2507.19510, 2025

    work page arXiv 2025