Pith. sign in

REVIEW 5 major objections 5 minor 24 references

Enhancing Multi-level Urban Instant Delivery Management via Infomap-based Hierarchical Community Detection

T0 review · 5 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read This paper claims that applying Infomap's hierarchical community detection to on-demand delivery networks in Shenzhen and Beijing reveals a stable three-level nesting of delivery clusters, and that this hierarchy can guide hub placement…

desk verdict Routine Infomap application to new delivery data with an unsupported optimization claim; the descriptive clusters are plausible, but the paper's central claim needs major revision or removal. read the letter →

arxiv 2411.11267 v2 pith:G4ZGOASV submitted 2024-11-18 physics.soc-ph

Chengbo Zhang This is my paper · ORCID
classification physics.soc-ph
keywords on-demanddeliveryhierarchicalcommunitydetectionInfomapmapequationspatialnetworkurbanlogisticsShenzhenBeijing
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper tries to show that hierarchical community detection, specifically Infomap's map-equation method, can uncover a nested three-level structure in the flow of on-demand delivery orders in two large Chinese cities. It builds weighted directed graphs from delivery data, with nodes as spatial grid units and edges as order counts, and runs Infomap to find communities at multiple scales. The reported result is that both Shenzhen and Beijing stabilize at three hierarchical levels, with Shenzhen showing five large regional clusters that subdivide into 30 and then 56 smaller zones, while Beijing shows 26, 92, and 123 clusters. The intended payoff is practical: if these clusters reflect real spatial dependencies, logistics planners could place hubs, assign routes, and allocate resources differently at each level. The paper argues the method is scalable and adaptable for data-driven urban logistics, though the optimization benefits are stated as implications rather than measured.

What carries the argument

The load-bearing object is Infomap's map equation $L(M)$, the expected description length of a random walk on the network when module boundaries are encoded with prefix and suffix codes. Infomap minimizes $L(M)$ to partition the network, then recurses inside each module to produce nested communities, with transition probabilities $P_{ij}=w_{ij}/\sum_{k}w_{ik}$ so that edge weights (order counts) directly shape the detected hierarchy. The depth_level parameter controls how many hierarchical levels are extracted, and the paper sets it to 3 because it reached stability at that value. The recursive minimization continues until no further improvement in $L(M)$ is possible, which is what makes the hierarchy emerge naturally from the flow structure.

What would settle it

Re-run Infomap on the same two datasets while sweeping depth_level from 1 to 5 and record the number of communities and the map-equation description length at each level. If the partition keeps changing beyond depth 3, or if a coarser or finer resolution of the spatial units changes the number of levels, the claim of a stable optimal three-level structure would be refuted.

Watch

Extended reading notes

Core claim

On its own terms, the paper's central claim is that the map-equation hierarchy of a directed, weighted delivery network is not flat: when Infomap is allowed to recurse, both Shenzhen and Beijing settle at a depth of three, and the resulting nested communities differ between the cities in a way that mirrors their urban form. Shenzhen's Level 1 has 5 broad communities, Beijing's has 26; the finer levels also fragment differently, with Beijing producing more numerous, smaller communities. The paper interprets this as evidence that delivery flows are organized in a nested spatial hierarchy, and that this hierarchy can be used to coordinate macro-level regional hubs with micro-level last-mile delivery zones.

Load-bearing premise

The whole three-level result rests on the assumption that the parameter controlling how many levels to extract is best set to three because it 'reached stability' at that value, but the stability criterion is never defined and no sensitivity analysis is shown.

Editorial extensions

If this is right

  • If the three-level structure is real, delivery management can be organized at three matching scales: broad regional flow regions, intermediate urban sub-regions, and fine neighborhood zones, with different route and resource strategies at each level.
  • The different cluster counts and sizes in Shenzhen versus Beijing imply that delivery-network planning should be city-specific rather than one-size-fits-all; Beijing's finer fragmentation suggests more localized last-mile management, while Shenzhen's broad clusters support hub-based regional coordination.
  • The method provides a data-driven way to define hierarchical service areas directly from order flows, which could complement traditional administrative boundaries or shipping-zone maps.
  • Since Infomap handles large directed and weighted networks efficiently, the same workflow can be rerun as orders accumulate, making the detected hierarchy a candidate basis for adaptive resource allocation over time.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The paper's 'optimization' conclusion is not tested: no delivery-cost or travel-time metric is computed before and after using the hierarchy. A direct test would assign orders to clusters and compare intra-cluster versus cross-cluster route lengths or courier idle time against a random or administrative partition.
  • The two datasets use different spatial units (about 1 km Voronoi polygons in Shenzhen, 500 m grids in Beijing), so the cross-city difference in cluster counts could partly reflect resolution rather than urban structure. Controlling spatial resolution would test that.
  • The depth_level 'stability' criterion is never defined; one could scan depth levels 2 through 5 with a defined resolution measure and see whether the three-level structure persists.
  • The hierarchy could be connected to temporal demand rhythms: detecting communities separately for peak and off-peak hours might reveal whether the three-level structure is stable in time or shifts with delivery demand.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

5 major / 5 minor

Summary. The paper applies Infomap-based hierarchical community detection to on-demand delivery flow networks constructed from Ele.me datasets for Shenzhen and Beijing. It reports a three-level community hierarchy for each city, discusses the cluster sizes and nesting relationships, and claims that this framework supports multi-level optimization of resource allocation and delivery strategies. The experiments are descriptive: they report community counts, average grid units per community, and Sankey/visual maps, but no operational performance metric is evaluated.

Significance. If the descriptive analysis is sound, the paper offers a potentially useful exploratory tool for urban logistics planning by revealing spatial hierarchies of delivery demand. The use of two large real-world datasets is a strength, and the hierarchical visualization is a clear way to communicate urban structure. However, the central claimed contribution—optimization of delivery management—is not demonstrated by the experiments, and key methodological details (depth_level stability, correct map equation, robustness) are missing. The paper's value as a descriptive case study is real but modest; its value as an optimization framework is unsubstantiated.

major comments (5)
  1. [Section IV-B] The claim that 'the depth_level parameter reached stability at a value of 3' is not substantiated: no stability criterion is defined, and no sensitivity analysis is provided. Because depth_level is a user-selected parameter that controls the number of hierarchical levels, the reported three-level structure for both cities may be an artifact of this choice rather than an intrinsic property of the networks. Please define the stability measure, report its values across a range of depth_level, and show that the cluster counts and memberships are robust to reasonable variations.
  2. [Abstract and Section V] The abstract and Section V state that the framework 'optimizes resource allocation and delivery strategies,' but Section IV contains no optimization step and no delivery-performance metric (e.g., delivery time, distance, order-to-rider allocation). The experiments report only community counts and average sizes. To support the central claim, the authors should either (i) add an evaluation showing that community-based routing or dispatch improves a concrete operational objective over baselines (e.g., administrative districts or flat partitions), or (ii) reframe the contribution as a descriptive/exploratory analysis of hierarchical delivery structure. As written, the optimization claim is unsupported.
  3. [Section III-B, Eq. (2)] The map equation in Eq. (2) is garbled: the formula as typeset is unreadable, and the symbols q, H, p, and ζ are not defined in the surrounding text. Since the entire method rests on minimizing this description length, please provide the correct equation (e.g., L(M) = q H(Q) + Σ_i p_i H(P_i)) and define every term. Without this, the method cannot be reproduced or checked.
  4. [Section IV-A] The Shenzhen and Beijing networks are constructed with different spatial resolutions (approximately 1 km Voronoi polygons versus 500 m grid cells) and different edge definitions. The comparison of 'cohesive' (Shenzhen) versus 'fragmented' (Beijing) community structure may therefore reflect the underlying spatial resolution rather than true urban differences. Please either use a common spatial unit or include a resolution-sensitivity analysis, and temper the cross-city comparison accordingly.
  5. [Section IV] Infomap's search can depend on random seeds, but the paper does not report results from multiple runs. The community counts and memberships at each level may be a single draw from a stochastic algorithm. Please report results from repeated runs (e.g., mean and standard deviation of community counts, or a consensus partition) to establish that the reported structure is stable.
minor comments (5)
  1. [Section III-B] The section title 'Abbreviations and Acronyms' is misleading; the section actually describes the Infomap algorithm and its steps. Consider retitling it to 'Infomap Overview' or similar.
  2. [Table I] The row containing '√√ × × ×' is ambiguous because it is not clear which columns the symbols correspond to. Please replace the symbols with explicit 'yes/no' entries for each algorithm.
  3. [Abstract] The sentence 'The experiment was conducted to large scale on-demand delivery datasets' is grammatically incorrect; it should be 'conducted on large-scale on-demand delivery datasets.'
  4. [Section V] The discussion of future work would benefit from an explicit acknowledgment of the limitations of the current static analysis, such as the absence of temporal dynamics and the lack of validation against operational outcomes.
  5. [References] Reference [24] is missing volume, page, and year information; please complete the bibliographic details.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: the Infomap hierarchy is an external-algorithm output, but the fixed depth_level=3 'optimal' claim and the unmeasured 'optimization' wording are validation and overreach issues, not demonstrated circular reductions.

full rationale

The paper's derivation chain is self-contained in the narrow sense needed for circularity analysis: edges are order counts between grid cells, Infomap minimizes the map equation (Eq. 2) over the observed network, and the reported communities are outputs of that external algorithm rather than values fitted to the paper's conclusions. None of the paper's equations is equivalent to its inputs by construction, and no parameter estimated from a subset of the data is later renamed as a prediction. The only self-citation, reference [9] (Z. Chengbo, a co-author), is used to support the background sentence 'As freight interactions become increasingly networked [9]' and is not load-bearing for any result, so it does not raise the score. Two concerns are real but do not constitute circularity. First, Section IV-B states that 'the depth_level parameter reached stability at a value of 3, indicating that the optimal community structure for both cities is defined by three hierarchical levels,' but the stability criterion is never defined and no sensitivity analysis is shown; if stability were defined as 'the output has three levels,' the conclusion would become tautological. As reported, this is an underspecified modeling choice that limits reproducibility, not a demonstrated reduction of the result to the input. Second, the abstract and conclusion claim the framework optimizes 'resource allocation and delivery strategies,' yet Section IV contains no delivery-performance metric, baseline, or optimization step; this is an unsupported external-validity claim, not a circular derivation. Because the central hierarchy is algorithmically generated and independently checkable, the paper receives a low circularity score of 2, reflecting only the minor self-citation and the parameter-selection ambiguity.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

The central claim rests on an external algorithm (Infomap), a graph abstraction, and a researcher-chosen depth parameter. No new entities are introduced.

free parameters (2)
  • depth_level = 3
    Controls the number of hierarchical levels in Infomap; the authors state it 'reached stability' at 3, but the stability criterion is not defined, making it a researcher-selected value that directly determines the reported three-level structure.
  • spatial resolution = Shenzhen: 0.001 deg (~1 km), Beijing: 500 m grids
    The choice of spatial aggregation units affects which clusters are detected; no sensitivity analysis is provided.
assumptions (4)
  • domain assumption Aggregating delivery orders into flows between grid cells (or Voronoi polygons) preserves the functional structure of urban delivery networks.
    Section III-A models the network as weighted directed edges between grid units; if this aggregation discards relevant routing information, the detected clusters may not correspond to operational delivery zones.
  • domain assumption Infomap's map equation correctly identifies meaningful hierarchical communities in urban delivery flow networks.
    The paper relies on Rosvall and Bergstrom's Infomap (ref [16]) as an established method; the map equation is taken as a valid objective function without validation on this specific network type.
  • domain assumption Recursively applying Infomap inside each community produces a meaningful nested hierarchy rather than an artifact of the algorithm.
    Section III-C assumes the recursive subdivision captures real multi-level spatial organization, but no comparison with alternative hierarchies or null models is given.
  • ad hoc to paper The depth_level parameter 'reached stability' at 3, implying an optimal number of levels.
    Section IV-B states stability without defining the metric; this assumption is specific to this paper and not independently justified.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Enhancing Multi-level Urban Instant Delivery Management via Infomap-based Hierarchical Community Detection." pith.science (2026). https://pith.science/paper/G4ZGOASV

@misc{pith2026241111267,
  author       = {Pith},
  title        = {Pith review of: Enhancing Multi-level Urban Instant Delivery Management via Infomap-based Hierarchical Community Detection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/G4ZGOASV}},
  note         = {Machine review of arXiv:2411.11267}
}
read the original abstract

Efficient management of on-demand delivery systems is essential for modern urban logistics, especially in densely populated cities with complex spatial layouts. This study introduces a novel, computer-supported cooperative framework that utilizes Infomap-based hierarchical community detection to analyze spatial multilevel clustering patterns. The experiment was conducted to large scale on-demand delivery datasets from Shenzhen and Beijing, revealing integrated spatial clusters that align with cohesive urban layout. Through hierarchical detection, finer and fragmented clusters are identified, reflecting its diverse urban structure and delivery demands. The findings demonstrate the effectiveness of hierarchical community detection in uncovering spatial dependencies and optimizing resource allocation and delivery strategies. This framework provides practical insights for urban logistics, enabling tailored approaches for business hub placement, route allocation, and adaptive resource management.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

24 extracted references · 12 canonical work pages

  1. [1]

    Integrating collection-and-deliverypointsinthestrategicdesignofurbanlast- milee-commercedistributionnetworks,

    M.Janjevic,M.Winkenbach,andD.Merchán,“Integrating collection-and-deliverypointsinthestrategicdesignofurbanlast- milee-commercedistributionnetworks,”Transp.Res.PartELogist. Transp.Rev.,vol.131,pp.37–67,Nov.2019,doi: 10.1016/j.tre.2019.09.001

    work page doi:10.1016/j.tre.2019.09.001 2019

  2. [2]

    Monitoringurbanaccessibility forfreightdeliveryservicesfromvehiclestracesandnetwork modelling,

    M.Pirra,A.Carboni,andF.Deflorio,“Monitoringurbanaccessibility forfreightdeliveryservicesfromvehiclestracesandnetwork modelling,”Transp.Res.Procedia,vol.41,pp.410–413,Jan.2019, doi:10.1016/j.trpro.2019.09.067

    work page doi:10.1016/j.trpro.2019.09.067 2019

  3. [3]

    Researchondeliveryproblembasedon two-stagemulti-objectiveoptimizationfortakeoutriders,

    C.Tang,C.Liu,andC.Li,“Researchondeliveryproblembasedon two-stagemulti-objectiveoptimizationfortakeoutriders,”J.Ind. Manag.Optim.,vol.19,no.11,pp.7881–7919,Nov.2023,doi: 10.3934/jimo.2023025

    work page doi:10.3934/jimo.2023025 2023

  4. [4]

    Theinfluenceofonlinefood deliveryservicesoneating-outtravel—Acasestudyofchengdu, China,

    E.Li,K.Shi,Y.Yang,andF.Witlox,“Theinfluenceofonlinefood deliveryservicesoneating-outtravel—Acasestudyofchengdu, China,”J.Retail.Consum.Serv.,vol.81,p.104023,Nov.2024,doi: 10.1016/j.jretconser.2024.104023

    work page arXiv 2024

  5. [5]

    Affectiveplatformurbanism:Changinghabitsofdigital on-demandconsumption,

    D.Bissell,“Affectiveplatformurbanism:Changinghabitsofdigital on-demandconsumption,”Geoforum,vol.115,pp.102–110,Oct. 2020,doi:10.1016/j.geoforum.2020.06.026

    work page doi:10.1016/j.geoforum.2020.06.026 2020

  6. [6]

    Communitylogistics fordynamicvehicledispatching:Theeffectsofcommunitydeparture ‘time’and‘space,’

    Z.Ouyang,E.K.H.Leung,andG.Q.Huang,“Communitylogistics fordynamicvehicledispatching:Theeffectsofcommunitydeparture ‘time’and‘space,’”Transp.Res.PartELogist.Transp.Rev.,vol.165, p.102842,Sep.2022,doi:10.1016/j.tre.2022.102842

    arXiv 2022

  7. [7]

    Communitylogistics anddynamiccommunitypartitioning:Anewapproachforsolvinge- commercelastmiledelivery,

    Z.Ouyang,E.K.H.Leung,andG.Q.Huang,“Communitylogistics anddynamiccommunitypartitioning:Anewapproachforsolvinge- commercelastmiledelivery,”Eur.J.Oper.Res.,vol.307,no.1,pp. 140–156,May2023,doi:10.1016/j.ejor.2022.08.029

    work page doi:10.1016/j.ejor.2022.08.029 2022

  8. [8]

    Findingcommunitystructureinspatially constrainedcomplexnetworks,

    Y.Chen,J.Xu,andM.Xu,“Findingcommunitystructureinspatially constrainedcomplexnetworks,”Int.J.Geogr.Inf.Sci.,vol.29,no.6, pp.889–911,Jun.2015,doi:10.1080/13658816.2014.999244

    arXiv 2015

Show all 24 references
  1. [9]

    Impactofspatialfunction complementarityonoutshoppingflows:Aspatialinteractionmodel,

    X.Zuopeng,L.Yonglin,andZ.Chengbo,“Impactofspatialfunction complementarityonoutshoppingflows:Aspatialinteractionmodel,” TravelBehav.Soc.,vol.39,p.100965,Apr.2025,doi: 10.1016/j.tbs.2024.100965

    2025

  2. [10]

    Uncoveringandmodelingthehierarchical organizationofurbanheavytruckflows,

    Y.Yangetal.,“Uncoveringandmodelingthehierarchical organizationofurbanheavytruckflows,”Transp.Res.PartELogist. Transp.Rev.,vol.179,p.103318,Nov.2023,doi: 10.1016/j.tre.2023.103318

    2023

  3. [11]

    Thetravelpatterndifference indocklessmicro-mobility:Sharede-bikesversussharedbikes,

    Q.Li,E.Zhang,D.Luca,andF.Fuerst,“Thetravelpatterndifference indocklessmicro-mobility:Sharede-bikesversussharedbikes,” Transp.Res.PartTransp.Environ.,vol.130,p.104179,May2024, doi:10.1016/j.trd.2024.104179

    2024

  4. [12]

    Dynamiccommunity detectionconsideringdailyrhythmsofhumanmobility,

    Y.Zhao,B.Y.Chen,F.Gao,andX.Zhu,“Dynamiccommunity detectionconsideringdailyrhythmsofhumanmobility,”Travel Behav.Soc.,vol.31,pp.209–222,Apr.2023,doi: 10.1016/j.tbs.2022.12.009

    2023 doi

  5. [13]

    Community- detection-basedspatialrangeidentificationforassessingbilateral jobs-housingbalance:ThecaseofBeijing,

    J.Wang,C.Zhou,J.Rong,S.Liu,andY.Wang,“Community- detection-basedspatialrangeidentificationforassessingbilateral jobs-housingbalance:ThecaseofBeijing,”Sustain.CitiesSoc.,vol. 87,p.104179,Dec.2022,doi:10.1016/j.scs.2022.104179

    2022

  6. [14]

    Identificationofcommunitiesinurban mobilitynetworksusingmulti-layergraphsofnetworktraffic,

    M.YildirimogluandJ.Kim,“Identificationofcommunitiesinurban mobilitynetworksusingmulti-layergraphsofnetworktraffic,” Transp.Res.PartCEmerg.Technol.,vol.89,pp.254–267,Apr.2018, doi:10.1016/j.trc.2018.02.015

    2018 doi

  7. [15]

    UnderstandingtheSpatialandTemporal ActivityPatternsofSubwayMobilityFlows,

    Z.Du,B.Yang,andJ.Liu,“UnderstandingtheSpatialandTemporal ActivityPatternsofSubwayMobilityFlows,”Feb.19,2017,arXiv: arXiv:1702.02456.Accessed:Jan.11,2024.[Online].Available: http://arxiv.org/abs/1702.02456

    2017 arXiv

  8. [16]

    Mapsofrandomwalksoncomplex networksrevealcommunitystructure,

    M.RosvallandC.T.Bergstrom,“Mapsofrandomwalksoncomplex networksrevealcommunitystructure,”Proc.Natl.Acad.Sci.,vol. 105,no.4,pp.1118–1123,Jan.2008,doi:10.1073/pnas.0706851105

    2008 doi

  9. [17]

    TheMealDeliveryRoutingProblem,

    A.L.Erera,D.Reyes,M.Savelsbergh,R.J.O’Neil,andS. Sahasrabudhe,“TheMealDeliveryRoutingProblem,”Apr.10,2018, OptimizationOnline:15139.Accessed:Mar.01,2025.[Online]. Available:https://optimization-online.org/?p=15139

    2018

  10. [18]

    Optimizationofcustomerservice anddriverdispatchareasforon-demandfooddelivery,

    J.Yang,H.C.Lau,andH.Wang,“Optimizationofcustomerservice anddriverdispatchareasforon-demandfooddelivery,”Transp.Res. PartCEmerg.Technol.,vol.165,p.104653,Aug.2024,doi: 10.1016/j.trc.2024.104653

    2024

  11. [19]

    ProvablyHigh-QualitySolutionsfor theMealDeliveryRoutingProblem,

    B.YildizandM.Savelsbergh,“ProvablyHigh-QualitySolutionsfor theMealDeliveryRoutingProblem,”Transp.Sci.,vol.53,no.5,pp. 1372–1388,Sep.2019,doi:10.1287/trsc.2018.0887

    2019

  12. [20]

    AutonomousDeliverySolutionsforLast-MileLogisticsOperations: ALiteratureReviewandResearchAgenda,

    V.Engesser,E.Rombaut,L.Vanhaverbeke,andP.Lebeau, “AutonomousDeliverySolutionsforLast-MileLogisticsOperations: ALiteratureReviewandResearchAgenda,”Sustainability,vol.15, no.3,Art.no.3,Jan.2023,doi:10.3390/su15032774

    2023 doi

  13. [21]

    Hierarchicalcommunitydetectionand functionalareaidentificationwithOSMroadsandcomplexgraph theory,

    Y.HongandY.Yao,“Hierarchicalcommunitydetectionand functionalareaidentificationwithOSMroadsandcomplexgraph theory,”Int.J.Geogr.Inf.Sci.,vol.33,no.8,pp.1569–1587,Aug. 2019,doi:10.1080/13658816.2019.1584806

    2019

  14. [22]

    Delineatingahierarchicalorganizationof rankedurbanclustersusingaspatialinteractionnetwork,

    T.Jia,X.Luo,andX.Li,“Delineatingahierarchicalorganizationof rankedurbanclustersusingaspatialinteractionnetwork,”Comput. Environ.UrbanSyst.,vol.87,p.101617,May2021,doi: 10.1016/j.compenvurbsys.2021.101617

    2021

  15. [23]

    ACity-WideCrowdsourcingDeliverySystemwith ReinforcementLearning,

    Y.Dingetal.,“ACity-WideCrowdsourcingDeliverySystemwith ReinforcementLearning,”Proc.ACMInteract.Mob.Wearable UbiquitousTechnol.,vol.5,no.3,pp.1–22,Sep.2021,doi: 10.1145/3478117

    2021 doi

  16. [24]

    Learning spatialinteractionrepresentationwithheterogeneousgraph convolutionalnetworksforurbanland-useinference,

    Z.Gong,C.Wang,Y.Chen,B.Liu,P.Zhao,andZ.Zhou,“Learning spatialinteractionrepresentationwithheterogeneousgraph convolutionalnetworksforurbanland-useinference,”Int.J.Geogr. Inf.Sci.,vol.0,no.0,pp.1–37,doi:10.1080/13658816.2024.2379473

    2024

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.