arxiv: 2605.02150 · v1 · submitted 2026-05-04 · 💻 cs.SI · cs.LG

Recognition: unknown

H3: A Healthcare Three-Hop Index for Physician Referral Network Prediction

Can Chen, Jiaxin Ying, Xu-Wen Wang, Zhexi Gu

Authors on Pith no claims yet

Pith reviewed 2026-05-08 02:28 UTC · model grok-4.3

classification 💻 cs.SI cs.LG

keywords physician referral networkslink predictionthree-hop indexhealthcare networksnetwork sparsityinterpretable modelsMedicare datacare coordination

0 comments

The pith

A three-hop index predicts missing physician referral links more accurately than heuristics or neural networks by tracing indirect paths with normalization and a penalty for redundancy.

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

The paper aims to show that physician referral networks have distinctive structure, such as sparsity and hub dominance, that standard link-prediction methods miss. H3 addresses this by scoring potential referral links according to the number of three-hop paths connecting two physicians, normalized by their degrees and reduced when paths share redundant intermediaries. Tested on Medicare shared-patient data, the index recovers both current links under sparse conditions and future links after time shifts, while also yielding explanations that point to the specific mediating physicians. If correct, this supplies a transparent, deployable tool for completing referral networks that supports care coordination without requiring opaque models.

Core claim

H3 models indirect referral pathways through intermediate physicians, with degree-based normalization and a redundancy penalty to mitigate hub-mediated noise. On Medicare Physician Shared Patient Patterns data, it outperforms classical heuristics and deep learning baselines in both within-period recovery of sparse links and cross-period prediction under temporal shift, while remaining fully decomposable to specific intermediary physicians.

What carries the argument

The H3 index, which aggregates three-hop paths between physicians while normalizing by degree and penalizing redundant intermediaries to score potential direct referral links.

Load-bearing premise

Medicare shared-patient patterns accurately reflect genuine referral relationships and three-hop paths are sufficient to represent the main referral mechanisms.

What would settle it

A test on a separate dataset of verified referral links in which H3 scores lower than the best baseline or in which removing the degree normalization or redundancy penalty improves accuracy.

Figures

Figures reproduced from arXiv: 2605.02150 by Can Chen, Jiaxin Ying, Xu-Wen Wang, Zhexi Gu.

**Figure 1.** Figure 1: Illustration of the H3 scoring framework. (A) Two-hop methods fail when vi and vj share no common neighbor, missing clinically meaningful indirect referral links. (B) L3 extends to three-hop paths but applies a fixed hub penalty (β = 0.5), insufficient for the heavy-tailed degree distribution of physician referral networks. (C) H3 scores each node pair by aggregating weighted three-hop paths vi → vk → vl →… view at source ↗

**Figure 2.** Figure 2: Link prediction performance across Task A (within-period) and Task B (cross-period). Best results are bold; second-best are underlined. TABLE II OVERVIEW OF EVALUATION TASKS FOR TEMPORAL LINK PREDICTION. Task Training Network Positive Test Links Within-period (Task A) Same calendar year; 50% links Held-out 50% from same snapshot Cross-period (Task B) 30-day short-window network New links in 90/180- day win… view at source ↗

**Figure 3.** Figure 3: Hyperparameter sensitivity of H3 grouped by design component. Each panel (G1–G4) corresponds to a parameter family defined in Table IV view at source ↗

**Figure 4.** Figure 4: AUPRC across link expansion ratio quartiles r under Task B. Each box shows the 5–95 percentile range over states. hypothesis that suppressing redundant intermediate connectors stabilizes ranking under sparse and noisy conditions. E. Robustness to Physician Mobility Expanding the observation window from a short snapshot to a longer period inherently increases incidental patient mobility and hub mixing, ampl… view at source ↗

read the original abstract

Accurate prediction of physician referral links is essential for optimizing care coordination and reducing fragmentation in healthcare delivery. However, existing computational methods, ranging from triadic closure heuristics to graph neural networks, fail to capture the intrinsic properties of physician referral networks, including sparsity, disassortative degree mixing, and hub-dominated topology. Here, we propose H3, a healthcare three-hop index that addresses these limitations by modeling indirect referral pathways through intermediate physicians, with degree-based normalization and a redundancy penalty to mitigate hub-mediated noise. Using Medicare Physician Shared Patient Patterns data, we evaluate H3 under two complementary prediction regimes: within-period prediction, which assesses recovery of contemporaneous referral links under sparse conditions, and cross-period prediction, which tests robustness to temporal shift as referral windows expand. Across both regimes, H3 consistently outperforms classical heuristics and deep learning-based baselines. Unlike black-box neural network approaches, H3 produces fully decomposable predictions traceable to specific intermediary physicians, offering a transparent and deployable solution for referral network completion.

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.

Desk Editor's Note private letter to a colleague

H3 is a domain-specific three-hop index with degree normalization and redundancy penalty for physician referral prediction that claims to beat heuristics and GNNs on Medicare data while staying interpretable, but the abstract shows no equations or numbers and the data proxy looks shaky.

read the letter

The main thing to know is that the authors built H3 as a simple counting index over three-hop paths in physician networks, with degree-based normalization to handle hubs and a penalty to cut redundant routes, and they report it beats both classical link-prediction heuristics and graph neural nets in two settings: recovering links inside the same time window and predicting forward across expanding referral periods on Medicare shared-patient data. The interpretability angle is the clearest practical win, since predictions can be traced back to specific intermediate physicians rather than sitting inside a black box.

Referee Report

2 major / 2 minor

Summary. The manuscript proposes H3, a three-hop index for physician referral network prediction that incorporates degree-based normalization and a redundancy penalty to address sparsity, disassortative mixing, and hub-dominated structure in referral graphs. Using Medicare Physician Shared Patient Patterns data, it evaluates H3 in within-period (contemporaneous link recovery) and cross-period (temporal robustness) regimes, claiming consistent outperformance over classical heuristics and deep learning baselines while providing interpretable, decomposable predictions traceable to specific intermediaries.

Significance. If the empirical superiority holds and the data proxy is appropriate, H3 offers a transparent, non-black-box alternative for link prediction that directly exploits observable network properties; this could support practical deployment in care coordination tools. The explicit interpretability via intermediary tracing is a notable strength relative to GNN baselines.

major comments (2)

[Data section] Data section: the central claim of outperformance rests on Medicare shared-patient patterns serving as a faithful proxy for true referral links, yet no independent validation (e.g., against verified referral records or external datasets) is provided; H3 is explicitly engineered around the observed sparsity, disassortativity, and hub structure of this specific dataset, so any non-referral artifacts (geography, hospital affiliation) would render the superiority an artifact rather than evidence of better referral modeling.
[Results section] Results section: the abstract asserts 'consistent outperformance' across regimes but the provided description contains no equations for the H3 index, no quantitative metrics (precision, recall, AUC with error bars), and no details on how the redundancy penalty is set or tuned; without these, the load-bearing empirical claim cannot be verified or reproduced.

minor comments (2)

[Abstract] Abstract: the phrase 'fully decomposable predictions traceable to specific intermediary physicians' is stated but not illustrated with an example decomposition or pseudocode in the main text.
[Introduction] Introduction: prior work on link prediction in healthcare networks (e.g., triadic closure applications) is referenced only generically; specific citations and direct comparisons of failure modes would strengthen the motivation.

Simulated Author's Rebuttal

2 responses · 1 unresolved

We thank the referee for their constructive feedback on the data proxy and empirical presentation. We address each major comment below, indicating planned revisions where appropriate.

read point-by-point responses

Referee: [Data section] Data section: the central claim of outperformance rests on Medicare shared-patient patterns serving as a faithful proxy for true referral links, yet no independent validation (e.g., against verified referral records or external datasets) is provided; H3 is explicitly engineered around the observed sparsity, disassortative mixing, and hub-dominated structure of this specific dataset, so any non-referral artifacts (geography, hospital affiliation) would render the superiority an artifact rather than evidence of better referral modeling.

Authors: We acknowledge the proxy nature of Medicare shared-patient data and agree that direct validation against verified referrals would strengthen claims. This proxy is standard in the literature (e.g., studies on physician networks by Barnett et al.), as it correlates strongly with referrals while respecting privacy constraints that preclude public verified records. The normalizations in H3 address general network properties (sparsity, disassortativity, hubs) documented across multiple referral datasets, not this one alone. We will add a limitations subsection with robustness checks for potential geographic/hospital artifacts. Partial revision: expanded discussion, but no new external validation data can be added. revision: partial
Referee: [Results section] Results section: the abstract asserts 'consistent outperformance' across regimes but the provided description contains no equations for the H3 index, no quantitative metrics (precision, recall, AUC with error bars), and no details on how the redundancy penalty is set or tuned; without these, the load-bearing empirical claim cannot be verified or reproduced.

Authors: The full manuscript defines H3 mathematically in Section 3, with the three-hop index formula, degree normalization, and redundancy penalty. Section 4 reports precision, recall, AUC with error bars over multiple splits, and tuning details (grid search on validation set with sensitivity analysis). We will revise to highlight the formulation earlier and add a summary table of all metrics for clarity and reproducibility. revision: yes

standing simulated objections not resolved

Independent validation of the Medicare shared-patient proxy against verified referral records (not publicly available due to privacy constraints)

Circularity Check

0 steps flagged

H3 index is an explicit heuristic formulation with no reduction to inputs by construction

full rationale

The paper defines H3 directly as a three-hop index incorporating degree-based normalization and a redundancy penalty to capture indirect referral pathways in sparse, hub-dominated networks. No equations or derivation steps are shown that reduce the index to a fitted parameter, self-cited uniqueness theorem, or input data property by construction. The two prediction regimes (within-period and cross-period) are standard evaluation splits on the Medicare dataset rather than tautological recoveries. No self-citation chains or ansatz smuggling are referenced for the core modeling choices. The outperformance claim rests on empirical comparison to baselines, which is independent of the index definition itself.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on the domain assumption that three-hop paths plus degree normalization capture referral behavior; no free parameters or invented entities are stated in the abstract.

axioms (1)

domain assumption Physician referral networks exhibit sparsity, disassortative degree mixing, and hub-dominated topology
Explicitly listed in the abstract as the intrinsic properties that existing methods fail to capture.

pith-pipeline@v0.9.0 · 5475 in / 1161 out tokens · 22573 ms · 2026-05-08T02:28:15.405736+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

55 extracted references · 4 canonical work pages · 2 internal anchors

[1]

Analysis of the us patient referral network,

C. An, A. J. O’Malley, D. N. Rockmore, and C. D. Stock, “Analysis of the us patient referral network,”Statistics in medicine, vol. 37, no. 5, pp. 847–866, 2018

2018
[2]

Patient referral patterns and the spread of hospital-acquired infections through national health care networks,

T. Donker, J. Wallinga, and H. Grundmann, “Patient referral patterns and the spread of hospital-acquired infections through national health care networks,”PLoS computational biology, vol. 6, no. 3, p. e1000715, 2010

2010
[3]

Effect of physician collaboration network on hospitalization cost and readmission rate,

S. Uddin, L. Hossain, and M. Kelaher, “Effect of physician collaboration network on hospitalization cost and readmission rate,”The European Journal of Public Health, vol. 22, no. 5, pp. 629–633, 2012

2012
[4]

Patient sharing among physicians and costs of care: a network analytic approach to care coordination using claims data,

C. E. Pollack, G. E. Weissman, K. W. Lemke, P. S. Hussey, and J. P. Weiner, “Patient sharing among physicians and costs of care: a network analytic approach to care coordination using claims data,”Journal of general internal medicine, vol. 28, no. 3, pp. 459–465, 2013

2013
[5]

Care patterns in medicare and their implications for pay for performance,

H. H. Pham, D. Schrag, A. S. O’Malley, B. Wu, and P. B. Bach, “Care patterns in medicare and their implications for pay for performance,” New England Journal of Medicine, vol. 356, no. 11, pp. 1130–1139, 2007

2007
[6]

Team rela- tionships and performance: Evidence from healthcare referral networks,

L. Agha, K. M. Ericson, K. H. Geissler, and J. B. Rebitzer, “Team rela- tionships and performance: Evidence from healthcare referral networks,” Management Science, vol. 68, no. 5, pp. 3735–3754, 2022

2022
[7]

Fragmented division of labor and healthcare costs: Evidence from moves across regions,

L. Agha, B. Frandsen, and J. B. Rebitzer, “Fragmented division of labor and healthcare costs: Evidence from moves across regions,”Journal of Public Economics, vol. 169, pp. 144–159, 2019

2019
[8]

Orga- nizational fragmentation and care quality in the us healthcare system,

R. D. Cebul, J. B. Rebitzer, L. J. Taylor, and M. E. V otruba, “Orga- nizational fragmentation and care quality in the us healthcare system,” Journal of Economic Perspectives, vol. 22, no. 4, pp. 93–113, 2008

2008
[9]

Referral paths in the us physician network,

C. An, A. J. O’Malley, and D. N. Rockmore, “Referral paths in the us physician network,”Applied network science, vol. 3, no. 1, p. 20, 2018

2018
[10]

Effects of incomplete inter- hospital network data on the assessment of transmission dynamics of hospital-acquired infections,

H. Xia, J. Horn, M. J. Piotrowska, K. Sakowski, A. Karch, H. Tahir, M. Kretzschmar, and R. Mikolajczyk, “Effects of incomplete inter- hospital network data on the assessment of transmission dynamics of hospital-acquired infections,”PLOS Computational Biology, vol. 17, no. 5, p. e1008941, 2021

2021
[11]

Physician patient-sharing networks and the cost and intensity of care in us hospitals,

M. L. Barnett, N. A. Christakis, J. O’Malley, J.-P. Onnela, N. L. Keating, and B. E. Landon, “Physician patient-sharing networks and the cost and intensity of care in us hospitals,”Medical care, vol. 50, no. 2, pp. 152–160, 2012

2012
[12]

Mapping physician networks with self-reported and administrative data,

M. L. Barnett, B. E. Landon, A. J. O’malley, N. L. Keating, and N. A. Christakis, “Mapping physician networks with self-reported and administrative data,”Health services research, vol. 46, no. 5, pp. 1592– 1609, 2011

2011
[13]

Variation in patient-sharing networks of physicians across the united states,

B. E. Landon, N. L. Keating, M. L. Barnett, J.-P. Onnela, S. Paul, A. J. O’Malley, T. Keegan, and N. A. Christakis, “Variation in patient-sharing networks of physicians across the united states,”Jama, vol. 308, no. 3, pp. 265–273, 2012

2012
[14]

Network interventions,

T. W. Valente, “Network interventions,”science, vol. 337, no. 6090, pp. 49–53, 2012

2012
[15]

Link prediction in complex networks: A survey,

L. Lü and T. Zhou, “Link prediction in complex networks: A survey,” Physica A: statistical mechanics and its applications, vol. 390, no. 6, pp. 1150–1170, 2011

2011
[16]

The link prediction problem for social networks,

D. Liben-Nowell and J. Kleinberg, “The link prediction problem for social networks,” inProceedings of the twelfth international conference on Information and knowledge management, 2003, pp. 556–559

2003
[17]

Friends and neighbors on the web,

L. A. Adamic and E. Adar, “Friends and neighbors on the web,”Social networks, vol. 25, no. 3, pp. 211–230, 2003

2003
[18]

Growing scale-free networks with tunable clustering,

P. Holme and B. J. Kim, “Growing scale-free networks with tunable clustering,”Physical review E, vol. 65, no. 2, p. 026107, 2002

2002
[19]

Predicting missing links via local information,

T. Zhou, L. Lü, and Y .-C. Zhang, “Predicting missing links via local information,”The European Physical Journal B, vol. 71, no. 4, pp. 623–630, 2009. 12

2009
[20]

Clustering and preferential attachment in growing networks,

M. E. Newman, “Clustering and preferential attachment in growing networks,”Physical review E, vol. 64, no. 2, p. 025102, 2001

2001
[21]

A comprehensive survey of link prediction methods: D. arrar et al

D. Arrar, N. Kamel, and A. Lakhfif, “A comprehensive survey of link prediction methods: D. arrar et al.”The journal of supercomputing, vol. 80, no. 3, pp. 3902–3942, 2024

2024
[22]

Link prediction in complex network using information flow,

F. Aziz, L. T. Slater, L. Bravo-Merodio, A. Acharjee, and G. V . Gkoutos, “Link prediction in complex network using information flow,”Scientific Reports, vol. 13, no. 1, p. 14660, 2023

2023
[23]

Link prediction on complex networks: an experimental survey,

H. Wu, C. Song, Y . Ge, and T. Ge, “Link prediction on complex networks: an experimental survey,”Data science and engineering, vol. 7, no. 3, pp. 253–278, 2022

2022
[24]

Network-based prediction of protein interactions,

I. A. Kovács, K. Luck, K. Spirohn, Y . Wang, C. Pollis, S. Schlabach, W. Bian, D.-K. Kim, N. Kishore, T. Haoet al., “Network-based prediction of protein interactions,”Nature communications, vol. 10, no. 1, p. 1240, 2019

2019
[25]

Experimental analyses on 2-hop- based and 3-hop-based link prediction algorithms,

T. Zhou, Y .-L. Lee, and G. Wang, “Experimental analyses on 2-hop- based and 3-hop-based link prediction algorithms,”Physica A: Statistical Mechanics and its Applications, vol. 564, p. 125532, 2021

2021
[26]

Medmetaverse: Medical care of chronic disease patients and managing data using artificial intelligence, blockchain, and wearable devices state-of-the-art methodology,

D. K. Murala, S. K. Panda, and S. P. Dash, “Medmetaverse: Medical care of chronic disease patients and managing data using artificial intelligence, blockchain, and wearable devices state-of-the-art methodology,”IEEE access, vol. 11, pp. 138 954–138 985, 2023

2023
[27]

Kipf et al.Variational Graph Auto- Encoders

T. N. Kipf and M. Welling, “Variational graph auto-encoders,”arXiv preprint arXiv:1611.07308, 2016

work page arXiv 2016
[28]

Semi-Supervised Classification with Graph Convolutional Networks

T. Kipf, “Semi-supervised classification with graph convolutional net- works,”arXiv preprint arXiv:1609.02907, 2016

work page internal anchor Pith review arXiv 2016
[29]

Inductive representation learning on large graphs,

W. Hamilton, Z. Ying, and J. Leskovec, “Inductive representation learning on large graphs,”Advances in neural information processing systems, vol. 30, 2017

2017
[30]

Graph attention networks,

P. Velickovic, G. Cucurull, A. Casanova, A. Romero, P. Lio, Y . Bengio et al., “Graph attention networks,”stat, vol. 1050, no. 20, pp. 10–48 550, 2017

2017
[31]

How Powerful are Graph Neural Networks?

K. Xu, W. Hu, J. Leskovec, and S. Jegelka, “How powerful are graph neural networks?”arXiv preprint arXiv:1810.00826, 2018

work page internal anchor Pith review arXiv 2018
[32]

Link prediction based on graph neural networks,

M. Zhang and Y . Chen, “Link prediction based on graph neural networks,” Advances in neural information processing systems, vol. 31, 2018

2018
[33]

Graph neural networks: link prediction,

M. Zhang, “Graph neural networks: link prediction,” inGraph Neural Networks: Foundations, Frontiers, and Applications. Springer, 2022, pp. 195–223

2022
[34]

Modeling polypharmacy side effects with graph convolutional networks,

M. Zitnik, M. Agrawal, and J. Leskovec, “Modeling polypharmacy side effects with graph convolutional networks,”Bioinformatics, vol. 34, no. 13, pp. i457–i466, 2018

2018
[35]

Graph representation learning in biomedicine and healthcare,

M. M. Li, K. Huang, and M. Zitnik, “Graph representation learning in biomedicine and healthcare,”Nature biomedical engineering, vol. 6, no. 12, pp. 1353–1369, 2022

2022
[36]

Stop explaining black box machine learning models for high stakes decisions and use interpretable models instead,

C. Rudin, “Stop explaining black box machine learning models for high stakes decisions and use interpretable models instead,”Nature machine intelligence, vol. 1, no. 5, pp. 206–215, 2019

2019
[37]

Improving clinical practice using clinical decision support systems: a systematic review of trials to identify features critical to success,

K. Kawamoto, C. A. Houlihan, E. A. Balas, and D. F. Lobach, “Improving clinical practice using clinical decision support systems: a systematic review of trials to identify features critical to success,”Bmj, vol. 330, no. 7494, p. 765, 2005

2005
[38]

Physician shared patient pat- terns data,

Centers for Medicare & Medicaid Services, “Physician shared patient pat- terns data,” https://www.cms.gov/Regulations-and-Guidance/Legislation/ FOIA/Referral-Data-FAQs, 2015, released via Freedom of Information Act request

2015
[39]

A review of link prediction algorithms in dynamic networks,

M. Sun and M. Tang, “A review of link prediction algorithms in dynamic networks,”Mathematics, vol. 13, no. 5, p. 807, 2025

2025
[40]

Evaluating link prediction methods,

Y . Yang, R. N. Lichtenwalter, and N. V . Chawla, “Evaluating link prediction methods,”Knowledge and Information Systems, vol. 45, no. 3, pp. 751–782, 2015

2015
[41]

Evaluating graph neural networks for link prediction: Current pitfalls and new benchmarking,

J. Li, H. Shomer, H. Mao, S. Zeng, Y . Ma, N. Shah, J. Tang, and D. Yin, “Evaluating graph neural networks for link prediction: Current pitfalls and new benchmarking,”Advances in Neural Information Processing Systems, vol. 36, pp. 3853–3866, 2023

2023
[42]

From link-prediction in brain connectomes and protein interactomes to the local-community- paradigm in complex networks,

C. V . Cannistraci, G. Alanis-Lobato, and T. Ravasi, “From link-prediction in brain connectomes and protein interactomes to the local-community- paradigm in complex networks,”Scientific reports, vol. 3, no. 1, p. 1613, 2013

2013
[43]

Deepwalk: Online learning of social representations,

B. Perozzi, R. Al-Rfou, and S. Skiena, “Deepwalk: Online learning of social representations,” inProceedings of the 20th ACM SIGKDD international conference on Knowledge discovery and data mining, 2014, pp. 701–710

2014
[44]

node2vec: Scalable feature learning for networks,

A. Grover and J. Leskovec, “node2vec: Scalable feature learning for networks,” inProceedings of the 22nd ACM SIGKDD international conference on Knowledge discovery and data mining, 2016, pp. 855– 864

2016
[45]

Weisfeiler and leman go neural: Higher-order graph neural networks,

C. Morris, M. Ritzert, M. Fey, W. L. Hamilton, J. E. Lenssen, G. Rattan, and M. Grohe, “Weisfeiler and leman go neural: Higher-order graph neural networks,” inProceedings of the AAAI conference on artificial intelligence, vol. 33, no. 01, 2019, pp. 4602–4609

2019
[46]

Class-imbalanced learning on graphs: A survey,

Y . Ma, Y . Tian, N. Moniz, and N. V . Chawla, “Class-imbalanced learning on graphs: A survey,”ACM Computing Surveys, vol. 57, no. 8, pp. 1–16, 2025

2025
[47]

Optimizing long-tailed link prediction in graph neural networks through structure representation enhancement,

Y . Wang, D. Wang, H. Liu, B. Hu, Y . Yan, Q. Zhang, and Z. Zhang, “Optimizing long-tailed link prediction in graph neural networks through structure representation enhancement,” inProceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, 2024, pp. 3222–3232

2024
[48]

The maximum capability of a topological feature in link prediction,

Y . Ran, X.-K. Xu, and T. Jia, “The maximum capability of a topological feature in link prediction,”PNAS nexus, vol. 3, no. 3, p. pgae113, 2024

2024
[49]

Elucidating the semantics-topology trade-off for knowledge inference-based pharma- cological discovery,

D. N. Sosa, G. Neculae, J. Fauqueur, and R. B. Altman, “Elucidating the semantics-topology trade-off for knowledge inference-based pharma- cological discovery,”Journal of Biomedical Semantics, vol. 15, no. 1, p. 5, 2024

2024
[50]

Open graph benchmark: Datasets for machine learning on graphs,

W. Hu, M. Fey, M. Zitnik, Y . Dong, H. Ren, B. Liu, M. Catasta, and J. Leskovec, “Open graph benchmark: Datasets for machine learning on graphs,”arXiv preprint arXiv:2005.00687, 2020

work page arXiv 2005
[51]

Assessment of community efforts to advance network-based prediction of protein–protein interactions,

X.-W. Wang, L. Madeddu, K. Spirohn, L. Martini, A. Fazzone, L. Bec- chetti, T. P. Wytock, I. A. Kovács, O. M. Balogh, B. Bencziket al., “Assessment of community efforts to advance network-based prediction of protein–protein interactions,”Nature communications, vol. 14, no. 1, p. 1582, 2023

2023
[52]

A survey of link prediction in temporal networks,

J. Xiong, A. Zareie, and R. Sakellariou, “A survey of link prediction in temporal networks,”SN Computer Science, vol. 7, no. 1, p. 100, 2026

2026
[53]

A survey of dynamic graph neural networks,

Y . Zheng, L. Yi, and Z. Wei, “A survey of dynamic graph neural networks,” Frontiers of Computer Science, vol. 19, no. 6, p. 196323, 2025

2025
[54]

Graph neural network modelling as a potentially effective method for predicting and analyzing procedures based on patients’ diagnoses,

J. G. D. Ochoa and F. E. Mustafa, “Graph neural network modelling as a potentially effective method for predicting and analyzing procedures based on patients’ diagnoses,”Artificial Intelligence in Medicine, vol. 131, p. 102359, 2022

2022
[55]

A systematic review of graph neural network in healthcare-based applications: Recent advances, trends, and future directions,

S. G. Paul, A. Saha, M. Z. Hasan, S. R. H. Noori, and A. Moustafa, “A systematic review of graph neural network in healthcare-based applications: Recent advances, trends, and future directions,”IEEE access, vol. 12, pp. 15 145–15 170, 2024. 13 TABLE VI PERFORMANCE COMPARISON ONOGBLINK PREDICTION BENCHMARKS.C= O G B L-C O L L A B (ACADEMIC COLLABORATION);C...

2024