Leveraging graph neural networks and mobility data for COVID-19 forecasting

Eduardo J. S. Luz; Fernando H. O. Duarte; Gladston J. P. Moreira; Leonardo B. L. Santos; Vander L. S. Freitas

arxiv: 2501.11711 · v3 · submitted 2025-01-20 · 💻 cs.LG · cs.SI

Leveraging graph neural networks and mobility data for COVID-19 forecasting

Fernando H. O. Duarte , Gladston J. P. Moreira , Eduardo J. S. Luz , Leonardo B. L. Santos , Vander L. S. Freitas This is my paper

Pith reviewed 2026-05-23 04:46 UTC · model grok-4.3

classification 💻 cs.LG cs.SI

keywords graph neural networksCOVID-19 forecastingmobility networksspatio-temporal modelsbackbone extractiondaily case predictionLSTM baseline

0 comments

The pith

Graph neural networks using sparsified mobility networks outperform LSTMs for daily COVID-19 case forecasts in Brazil and China.

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

The paper tests whether adding spatial structure from human mobility data improves forecasts over simple time-series models. It finds that GNN variants deliver lower error on volatile daily counts once the input graph is sparsified, while LSTMs remain competitive on smooth cumulative totals. The authors show that removing weak edges through backbone extraction stabilizes predictions and that temporal resolution matters for model choice. They also recast the task as binary classification to examine how context length interacts with forecast horizon. The work therefore isolates the conditions under which spatial dependencies become necessary rather than optional.

Core claim

Structural sparsification of mobility graphs combined with GNN architectures (GCRN and GCLSTM) yields statistically significant gains over LSTM baselines on daily case counts from Brazil and China, while the same architectures show no advantage on cumulative series; backbone extraction is presented as the key step that removes negligible connections and improves stability.

What carries the argument

Backbone extraction applied to human mobility networks to produce sparse graphs that are then fed to graph convolutional recurrent networks (GCRN) and graph convolutional LSTMs (GCLSTM) for joint spatial-temporal prediction.

If this is right

Spatial dependencies captured by GNNs become essential once daily rather than cumulative counts are the target.
Backbone extraction improves both accuracy and stability of spatio-temporal forecasts.
Temporal granularity of the input series determines whether GNNs or pure temporal baselines are preferable.
Framing epidemic forecasting as binary classification clarifies the trade-off between context size and prediction horizon.

Where Pith is reading between the lines

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

The same sparsification-plus-GNN pipeline could be tested on other mobility-linked outbreaks to check whether daily volatility is the general trigger for spatial modeling.
Real-time updates to the mobility backbone might allow the models to adapt faster to policy changes or new variants than static graphs permit.
If the binary-classification view holds, epidemic early-warning systems could be tuned by optimizing context length rather than by adding more complex architectures.

Load-bearing premise

The mobility networks accurately represent the spatial dependencies that drive transmission and backbone extraction discards only non-predictive edges.

What would settle it

Re-running the daily-case experiments on the same Brazil and China datasets but with the unsparsified full mobility graph or an alternative mobility source that produces no statistically significant GNN advantage over LSTM would falsify the central performance claim.

Figures

Figures reproduced from arXiv: 2501.11711 by Eduardo J. S. Luz, Fernando H. O. Duarte, Gladston J. P. Moreira, Leonardo B. L. Santos, Vander L. S. Freitas.

**Figure 2.** Figure 2: Data Preprocessing Workflow. Taking into account a temporal dataset that covers 1, 095 days, the division for Brazil results in 876 days allocated for training and 219 days for testing. For China, out of a total of 694 days, 555 are for training and 139 for testing. The application of sliding windows offers flexibility in the number of snapshots generated, which is which is adaptable to various experiment… view at source ↗

**Figure 3.** Figure 3: Experiment 4: Average RMSE Heatmaps for Regression Task, Brazil. [PITH_FULL_IMAGE:figures/full_fig_p015_3.png] view at source ↗

**Figure 4.** Figure 4: Experiment 5: Average RMSE Heatmaps for Regression Task, China. [PITH_FULL_IMAGE:figures/full_fig_p017_4.png] view at source ↗

**Figure 5.** Figure 5: Experiment 6: Average F1-Score Heatmaps for Classification Task, Brazil. [PITH_FULL_IMAGE:figures/full_fig_p018_5.png] view at source ↗

**Figure 6.** Figure 6: Experiment 6: Average Precision Heatmaps for Classification Task, Brazil. [PITH_FULL_IMAGE:figures/full_fig_p019_6.png] view at source ↗

**Figure 7.** Figure 7: Experiment 6: Average Recall Heatmaps for Classification Task, Brazil. [PITH_FULL_IMAGE:figures/full_fig_p019_7.png] view at source ↗

**Figure 8.** Figure 8: Average F1-Score Heatmaps for Classification Task, China. [PITH_FULL_IMAGE:figures/full_fig_p020_8.png] view at source ↗

**Figure 9.** Figure 9: Average Precision Heatmaps for Classification Task, China. [PITH_FULL_IMAGE:figures/full_fig_p021_9.png] view at source ↗

**Figure 10.** Figure 10: Average Recall Heatmaps for Classification Task, China. [PITH_FULL_IMAGE:figures/full_fig_p021_10.png] view at source ↗

read the original abstract

The COVID-19 pandemic has claimed millions of lives, spurring the development of diverse forecasting models. In this context, the true utility of complex spatio-temporal architectures versus simpler temporal baselines remains a subject of debate. Here, we show that structural sparsification of the input graph and temporal granularity are determining factors for the effectiveness of Graph Neural Networks (GNNs). By leveraging human mobility networks in Brazil and China, we address a conflicting scenario in the literature: while standard LSTMs suffice for smooth, monotonic cumulative trends, GNNs significantly outperform baselines when forecasting volatile daily case counts. We show that backbone extraction substantially enhances predictive stability and reduces predictive error by removing negligible connections. Our results indicate that incorporating spatial dependencies is essential for modeling complex dynamics. Specifically, GNN architectures such as GCRN and GCLSTM outperform the LSTM baseline (Nemenyi test, p < 0.05) on datasets from Brazil and China for daily case predictions. Lastly, we frame the problem as a binary classification task to better analyze the dependency between context sizes and prediction horizons.

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

GNNs beat LSTM on daily COVID forecasts with sparsified mobility graphs from Brazil and China, but the design skips the null-graph control needed to credit the mobility edges themselves.

read the letter

The paper's core finding is that GCRN and GCLSTM outperform a plain LSTM on daily case counts (Nemenyi p<0.05) once the mobility graph is sparsified by backbone extraction, while LSTM is adequate for smoother cumulative counts. They also recast the task as binary classification to probe context size versus horizon. That conditional result is the useful part: it gives practitioners a concrete rule of thumb about when spatial structure helps rather than claiming GNNs are always better. The sparsification observation is practical and worth noting for anyone building these models. The datasets from two countries add some geographic breadth to the existing literature on GNN epidemic forecasting. The main gap is exactly the one in the stress-test note. The authors attribute the lift to the mobility-derived spatial dependencies, yet they only compare against a non-graph LSTM. A control that swaps in a degree-matched random graph or other null adjacency while holding architecture, training, and granularity fixed is missing. Without it, the gains could trace to the GNN inductive bias, the sparsification step, or simply having any graph rather than the specific mobility edges. Dataset sizes, exact preprocessing steps, and error-bar reporting are also thin in the abstract, so the full text needs to supply those to make the statistical claim convincing. This work is aimed at applied modelers who already know the standard GNN time-series toolkit and want evidence on when to reach for it in public-health settings. It is not a methods advance, but the empirical question is real and the new regional data is a modest plus. I would send it to peer review with a request to add the null-graph ablation and clearer experimental details; the central claim is testable and the paper is coherent enough to benefit from referee input.

Referee Report

2 major / 2 minor

Summary. The manuscript claims that GNN architectures (GCRN, GCLSTM) outperform LSTM baselines for daily COVID-19 case forecasting on Brazil and China mobility networks (Nemenyi p<0.05), that backbone extraction via sparsification improves stability and reduces error by removing negligible connections, and that spatial dependencies encoded in the mobility graphs are essential for volatile daily counts (while LSTMs suffice for smooth cumulative trends). It additionally frames forecasting as binary classification to examine context-size vs. horizon dependencies.

Significance. If the results hold after appropriate controls, the work would clarify when and why GNNs add value over temporal baselines in epidemic forecasting and demonstrate the utility of mobility-graph sparsification, providing actionable guidance for spatio-temporal model selection in public-health applications.

major comments (2)

[Experimental results] The experimental comparison of GNNs to LSTM (reported in the results on daily-case predictions) does not include an ablation replacing the empirical mobility adjacency matrix with a degree-matched random graph or other null model while holding architecture, training, and temporal granularity fixed. This control is required to substantiate the claim that the specific mobility-derived structure (rather than GNN inductive bias or sparsification alone) drives the reported gains.
[Results on sparsification] The assertion that backbone extraction 'substantially enhances predictive stability and reduces predictive error' lacks quantification of the edge-removal fraction, before/after performance deltas with error bars, and a check that predictive signal is retained rather than discarded; these details are load-bearing for the sparsification claim.

minor comments (2)

[Methods] Dataset sizes, exact preprocessing steps, number of nodes/edges in the mobility networks, and sparsification threshold values should be stated explicitly (currently referenced only at a high level) to support reproducibility.
[Introduction] Ensure first-use definitions for all acronyms (GCRN, GCLSTM) and consistent notation for the binary-classification reformulation.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback. The suggested controls will strengthen the manuscript's claims about the role of mobility-graph structure and the benefits of sparsification. We address each major comment below and will incorporate the requested analyses.

read point-by-point responses

Referee: [Experimental results] The experimental comparison of GNNs to LSTM (reported in the results on daily-case predictions) does not include an ablation replacing the empirical mobility adjacency matrix with a degree-matched random graph or other null model while holding architecture, training, and temporal granularity fixed. This control is required to substantiate the claim that the specific mobility-derived structure (rather than GNN inductive bias or sparsification alone) drives the reported gains.

Authors: We agree that a null-model ablation is necessary to isolate the contribution of the empirical mobility structure. In the revised manuscript we will add this control: the empirical adjacency matrix will be replaced by a degree-matched random graph while keeping the GNN architecture (GCRN/GCLSTM), training procedure, temporal granularity, and sparsification (where applicable) identical. Results will be reported alongside the original experiments with the same statistical testing. revision: yes
Referee: [Results on sparsification] The assertion that backbone extraction 'substantially enhances predictive stability and reduces predictive error' lacks quantification of the edge-removal fraction, before/after performance deltas with error bars, and a check that predictive signal is retained rather than discarded; these details are load-bearing for the sparsification claim.

Authors: We acknowledge that the current manuscript provides insufficient quantitative detail on sparsification. The revision will report: (1) the precise fraction of edges removed by backbone extraction on each dataset, (2) before/after performance metrics (MAE, RMSE, and stability measures) with error bars across multiple random seeds, and (3) a control comparing backbone extraction against random edge removal at the same density to verify that predictive signal is retained rather than discarded. revision: yes

Circularity Check

0 steps flagged

No significant circularity; purely empirical evaluation

full rationale

The manuscript reports experimental comparisons of GCRN, GCLSTM, and LSTM models on Brazil/China mobility-augmented COVID-19 time series, with performance assessed via Nemenyi tests. No equations, derivations, fitted parameters renamed as predictions, uniqueness theorems, or ansatzes appear in the text. All claims rest on held-out forecasting metrics and statistical tests performed on external datasets, rendering the evaluation self-contained and independent of any definitional or self-citational reduction.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Only abstract available; no explicit free parameters, axioms, or invented entities are described.

pith-pipeline@v0.9.0 · 5741 in / 980 out tokens · 19698 ms · 2026-05-23T04:46:19.105607+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

28 extracted references · 28 canonical work pages · 1 internal anchor

[1]

V. L. S. Freitas, T. C. R. O. Konstantyner, J. F. Mendes, C. S. N. Sepetauskas, L. B. L. Santos, The correspondence between the struc- ture of the terrestrial mobility network and the spreading of covid-19 in brazil, Cadernos de Sa´ ude P´ ublica 36 (2020) e00184820

work page 2020
[2]

V. L. Freitas, G. J. Moreira, L. B. Santos, Robustness analysis in an inter-cities mobility network: modeling municipal, state and federal ini- tiatives as failures and attacks toward sars-cov-2 containment, PeerJ 8 (2020) e10287

work page 2020
[3]

Mu˜ noz-Organero, P

M. Mu˜ noz-Organero, P. Callejo, M.´A. Hombrados-Herrera, A new rnn based machine learning model to forecast covid-19 incidence, enhanced by the use of mobility data from the bike-sharing service in madrid, Heliyon 9 (6) (2023)

work page 2023
[4]

Witzke, N

S. Witzke, N. Danz, K. Baum, B. Y. Renard, Mobility data improve forecasting of covid-19 incidence trends using graph neural networks, in: epiDAMIK 6.0: The 6th International workshop on Epidemiology meets Data Mining and Knowledge Discovery at KDD 2023, 2023

work page 2023
[5]

Kapoor, X

A. Kapoor, X. Ben, L. Liu, B. Perozzi, M. Barnes, M. Blais, S. O’Banion, Examining covid-19 forecasting using spatio-temporal graph neural net- works, arXiv preprint arXiv:2007.03113 (2020)

work page arXiv 2007
[6]

Sarkar, A

S. Sarkar, A. Alhamadani, C.-T. Lu, Explainable prediction of the sever- ity of covid-19 outbreak for us counties, in: 2022 IEEE International Conference on Big Data (Big Data), IEEE, 2022, pp. 5338–5345

work page 2022
[7]

H. Li, R. Wei, W. Wang, N. Yu, Predicting covid-19 transmission in southern california with machine learning methods, in: 2024 9th Inter- national Conference on Big Data Analytics (ICBDA), IEEE, 2024, pp. 1–10. 24

work page 2024
[8]

F. H. O. Duarte, G. J. P. Moreira, E. J. S. Luz, L. B. L. Santos, V. L. S. Freitas, Time series forecasting of covid-19 cases in brazil with gnn and mobility networks, in: M. C. Naldi, R. A. C. Bianchi (Eds.), Intelligent Systems, Springer Nature Switzerland, Cham, 2023, pp. 361–375

work page 2023
[9]

Y. Seo, M. Defferrard, P. Vandergheynst, X. Bresson, Structured se- quence modeling with graph convolutional recurrent networks, in: Neu- ral Information Processing, 2018, pp. 362–373

work page 2018
[10]

J. Chen, X. Wang, X. Xu, Gc-lstm: graph convolution embedded lstm for dynamic network link prediction, Applied Intelligence 52 (7) (2022) 7513–7528

work page 2022
[11]

Comito, C

C. Comito, C. Pizzuti, Artificial intelligence for forecasting and diagnosing covid-19 pandemic: A focused review, Arti- ficial Intelligence in Medicine 128 (2022) 102286. doi:https: //doi.org/10.1016/j.artmed.2022.102286. URL https://www.sciencedirect.com/science/article/pii/ S0933365722000513

work page doi:10.1016/j.artmed.2022.102286 2022
[12]

R. P. Joshi, V. Pejaver, N. E. Hammarlund, H. Sung, S. K. Lee, A. Furmanchuk, H.-Y. Lee, G. Scott, S. Gombar, N. Shah, S. Shen, A. Nassiri, D. Schneider, F. S. Ahmad, D. Liebovitz, A. Kho, S. Mooney, B. A. Pinsky, N. Banaei, A predictive tool for identifica- tion of sars-cov-2 pcr-negative emergency department patients using routine test results, Journal ...

work page doi:10.1016/j.jcv.2020.104502 2020
[13]

Pirouz, S

B. Pirouz, S. Shaffiee Haghshenas, S. Shaffiee Haghshenas, P. Piro, Investigating a serious challenge in the sustainable development pro- cess: Analysis of confirmed cases of covid-19 (new type of coronavirus) through a binary classification using artificial intelligence and regression analysis, Sustainability 12 (6) (2020). doi:10.3390/su12062427. URL ht...

work page doi:10.3390/su12062427 2020
[14]

F. H. O. Duarte, G. J. Moreira, E. J. Luz, L. B. Santos, V. L. Freitas, Correlations between epidemiological time series forecasting and influence regions of brazilian cities, 2023, p. 363 – 368, cited by: 0. 25 URL https://www.scopus.com/inward/record. uri?eid=2-s2.0-85181098306&partnerID=40&md5= f60d73eee623d21c9b7119bbc4407319

work page 2023
[15]

M. R. Davahli, K. Fiok, W. Karwowski, A. M. Aljuaid, R. Taiar, Predict- ing the dynamics of the covid-19 pandemic in the united states using graph theory-based neural networks, International journal of environ- mental research and public health 18 (7) (2021) 3834

work page 2021
[16]

H. Xie, D. Li, Y. Wang, Y. Kawai, Visualization method for the spread- ing curve of covid-19 in universities using gnn, in: 2022 IEEE Inter- national Conference on Big Data and Smart Computing (BigComp), IEEE, 2022, pp. 121–128

work page 2022
[17]

Fanelli, F

D. Fanelli, F. Piazza, Analysis and forecast of covid-19 spreading in china, italy and france, Chaos, Solitons & Fractals 134 (2020) 109761

work page 2020
[18]

C.-P. Kuo, J. S. Fu, Evaluating the impact of mobility on covid-19 pandemic with machine learning hybrid predic- tions, Science of The Total Environment 758 (2021) 144151. doi:https://doi.org/10.1016/j.scitotenv.2020.144151. URL https://www.sciencedirect.com/science/article/pii/ S0048969720376828

work page doi:10.1016/j.scitotenv.2020.144151 2021
[19]

Menczer, S

F. Menczer, S. Fortunato, C. A. Davis, A First Course in Network Sci- ence, Cambridge University Press, 2020. doi:10.1017/9781108653947

work page doi:10.1017/9781108653947 2020
[20]

C. H. Gomes Ferreira, F. Murai, A. P. C. Silva, M. Trevisan, L. Vassio, I. Drago, M. Mellia, J. M. Almeida, On network backbone extraction for modeling online collective behavior, PLOS ONE 17 (9) (2022) 1–36. doi:10.1371/journal.pone.0274218. URL https://doi.org/10.1371/journal.pone.0274218

work page doi:10.1371/journal.pone.0274218 2022
[21]

Z. Wu, S. Pan, F. Chen, G. Long, C. Zhang, P. S. Yu, A comprehen- sive survey on graph neural networks, IEEE Transactions on Neural Networks and Learning Systems 32 (2021) 4–24. doi:10.1109/TNNLS. 2020.2978386

work page doi:10.1109/tnnls 2021
[22]

Niepert, M

M. Niepert, M. Ahmed, K. Kutzkov, Learning convolutional neural net- works for graphs, in: International conference on machine learning, PMLR, 2016, pp. 2014–2023. 26

work page 2016
[23]

T. N. Kipf, M. Welling, Semi-supervised classification with graph con- volutional networks, arXiv preprint arXiv:1609.02907 (2016)

work page internal anchor Pith review Pith/arXiv arXiv 2016
[24]

IBGE, Liga¸ c˜ oes rodovi´ arias e hidrovi´ arias: 2016, IBGE, Coordena¸ c˜ ao de Geografia Rio de Janeiro, Brazil, 2017

work page 2016
[25]

doi:10.7910/DVN/ FAEZIO

China Data Lab, Baidu Mobility Data (2020). doi:10.7910/DVN/ FAEZIO. URL https://doi.org/10.7910/DVN/FAEZIO

work page doi:10.7910/dvn/ 2020
[26]

Cota, Monitoring the number of covid-19 cases and deaths in brazil at municipal and federative units level, SciELO Preprints (2020)

W. Cota, Monitoring the number of covid-19 cases and deaths in brazil at municipal and federative units level, SciELO Preprints (2020)

work page 2020
[27]

Safra Center for Ethics, Key metrics for covid suppression a framework for policy makers and the public (2020)

Harvard Global Health Institute and Harvard’s Edmond J. Safra Center for Ethics, Key metrics for covid suppression a framework for policy makers and the public (2020)

work page 2020
[28]

M. Jin, H. Y. Koh, Q. Wen, D. Zambon, C. Alippi, G. I. Webb, I. King, S. Pan, A survey on graph neural networks for time series: Forecasting, classification, imputation, and anomaly detection, IEEE Transactions on Pattern Analysis and Machine Intelligence 46 (12) (2024) 10466–10485. doi:10.1109/TPAMI.2024.3443141. 27

work page doi:10.1109/tpami.2024.3443141 2024

[1] [1]

V. L. S. Freitas, T. C. R. O. Konstantyner, J. F. Mendes, C. S. N. Sepetauskas, L. B. L. Santos, The correspondence between the struc- ture of the terrestrial mobility network and the spreading of covid-19 in brazil, Cadernos de Sa´ ude P´ ublica 36 (2020) e00184820

work page 2020

[2] [2]

V. L. Freitas, G. J. Moreira, L. B. Santos, Robustness analysis in an inter-cities mobility network: modeling municipal, state and federal ini- tiatives as failures and attacks toward sars-cov-2 containment, PeerJ 8 (2020) e10287

work page 2020

[3] [3]

Mu˜ noz-Organero, P

M. Mu˜ noz-Organero, P. Callejo, M.´A. Hombrados-Herrera, A new rnn based machine learning model to forecast covid-19 incidence, enhanced by the use of mobility data from the bike-sharing service in madrid, Heliyon 9 (6) (2023)

work page 2023

[4] [4]

Witzke, N

S. Witzke, N. Danz, K. Baum, B. Y. Renard, Mobility data improve forecasting of covid-19 incidence trends using graph neural networks, in: epiDAMIK 6.0: The 6th International workshop on Epidemiology meets Data Mining and Knowledge Discovery at KDD 2023, 2023

work page 2023

[5] [5]

Kapoor, X

A. Kapoor, X. Ben, L. Liu, B. Perozzi, M. Barnes, M. Blais, S. O’Banion, Examining covid-19 forecasting using spatio-temporal graph neural net- works, arXiv preprint arXiv:2007.03113 (2020)

work page arXiv 2007

[6] [6]

Sarkar, A

S. Sarkar, A. Alhamadani, C.-T. Lu, Explainable prediction of the sever- ity of covid-19 outbreak for us counties, in: 2022 IEEE International Conference on Big Data (Big Data), IEEE, 2022, pp. 5338–5345

work page 2022

[7] [7]

H. Li, R. Wei, W. Wang, N. Yu, Predicting covid-19 transmission in southern california with machine learning methods, in: 2024 9th Inter- national Conference on Big Data Analytics (ICBDA), IEEE, 2024, pp. 1–10. 24

work page 2024

[8] [8]

F. H. O. Duarte, G. J. P. Moreira, E. J. S. Luz, L. B. L. Santos, V. L. S. Freitas, Time series forecasting of covid-19 cases in brazil with gnn and mobility networks, in: M. C. Naldi, R. A. C. Bianchi (Eds.), Intelligent Systems, Springer Nature Switzerland, Cham, 2023, pp. 361–375

work page 2023

[9] [9]

Y. Seo, M. Defferrard, P. Vandergheynst, X. Bresson, Structured se- quence modeling with graph convolutional recurrent networks, in: Neu- ral Information Processing, 2018, pp. 362–373

work page 2018

[10] [10]

J. Chen, X. Wang, X. Xu, Gc-lstm: graph convolution embedded lstm for dynamic network link prediction, Applied Intelligence 52 (7) (2022) 7513–7528

work page 2022

[11] [11]

Comito, C

C. Comito, C. Pizzuti, Artificial intelligence for forecasting and diagnosing covid-19 pandemic: A focused review, Arti- ficial Intelligence in Medicine 128 (2022) 102286. doi:https: //doi.org/10.1016/j.artmed.2022.102286. URL https://www.sciencedirect.com/science/article/pii/ S0933365722000513

work page doi:10.1016/j.artmed.2022.102286 2022

[12] [12]

R. P. Joshi, V. Pejaver, N. E. Hammarlund, H. Sung, S. K. Lee, A. Furmanchuk, H.-Y. Lee, G. Scott, S. Gombar, N. Shah, S. Shen, A. Nassiri, D. Schneider, F. S. Ahmad, D. Liebovitz, A. Kho, S. Mooney, B. A. Pinsky, N. Banaei, A predictive tool for identifica- tion of sars-cov-2 pcr-negative emergency department patients using routine test results, Journal ...

work page doi:10.1016/j.jcv.2020.104502 2020

[13] [13]

Pirouz, S

B. Pirouz, S. Shaffiee Haghshenas, S. Shaffiee Haghshenas, P. Piro, Investigating a serious challenge in the sustainable development pro- cess: Analysis of confirmed cases of covid-19 (new type of coronavirus) through a binary classification using artificial intelligence and regression analysis, Sustainability 12 (6) (2020). doi:10.3390/su12062427. URL ht...

work page doi:10.3390/su12062427 2020

[14] [14]

F. H. O. Duarte, G. J. Moreira, E. J. Luz, L. B. Santos, V. L. Freitas, Correlations between epidemiological time series forecasting and influence regions of brazilian cities, 2023, p. 363 – 368, cited by: 0. 25 URL https://www.scopus.com/inward/record. uri?eid=2-s2.0-85181098306&partnerID=40&md5= f60d73eee623d21c9b7119bbc4407319

work page 2023

[15] [15]

M. R. Davahli, K. Fiok, W. Karwowski, A. M. Aljuaid, R. Taiar, Predict- ing the dynamics of the covid-19 pandemic in the united states using graph theory-based neural networks, International journal of environ- mental research and public health 18 (7) (2021) 3834

work page 2021

[16] [16]

H. Xie, D. Li, Y. Wang, Y. Kawai, Visualization method for the spread- ing curve of covid-19 in universities using gnn, in: 2022 IEEE Inter- national Conference on Big Data and Smart Computing (BigComp), IEEE, 2022, pp. 121–128

work page 2022

[17] [17]

Fanelli, F

D. Fanelli, F. Piazza, Analysis and forecast of covid-19 spreading in china, italy and france, Chaos, Solitons & Fractals 134 (2020) 109761

work page 2020

[18] [18]

C.-P. Kuo, J. S. Fu, Evaluating the impact of mobility on covid-19 pandemic with machine learning hybrid predic- tions, Science of The Total Environment 758 (2021) 144151. doi:https://doi.org/10.1016/j.scitotenv.2020.144151. URL https://www.sciencedirect.com/science/article/pii/ S0048969720376828

work page doi:10.1016/j.scitotenv.2020.144151 2021

[19] [19]

Menczer, S

F. Menczer, S. Fortunato, C. A. Davis, A First Course in Network Sci- ence, Cambridge University Press, 2020. doi:10.1017/9781108653947

work page doi:10.1017/9781108653947 2020

[20] [20]

C. H. Gomes Ferreira, F. Murai, A. P. C. Silva, M. Trevisan, L. Vassio, I. Drago, M. Mellia, J. M. Almeida, On network backbone extraction for modeling online collective behavior, PLOS ONE 17 (9) (2022) 1–36. doi:10.1371/journal.pone.0274218. URL https://doi.org/10.1371/journal.pone.0274218

work page doi:10.1371/journal.pone.0274218 2022

[21] [21]

Z. Wu, S. Pan, F. Chen, G. Long, C. Zhang, P. S. Yu, A comprehen- sive survey on graph neural networks, IEEE Transactions on Neural Networks and Learning Systems 32 (2021) 4–24. doi:10.1109/TNNLS. 2020.2978386

work page doi:10.1109/tnnls 2021

[22] [22]

Niepert, M

M. Niepert, M. Ahmed, K. Kutzkov, Learning convolutional neural net- works for graphs, in: International conference on machine learning, PMLR, 2016, pp. 2014–2023. 26

work page 2016

[23] [23]

T. N. Kipf, M. Welling, Semi-supervised classification with graph con- volutional networks, arXiv preprint arXiv:1609.02907 (2016)

work page internal anchor Pith review Pith/arXiv arXiv 2016

[24] [24]

IBGE, Liga¸ c˜ oes rodovi´ arias e hidrovi´ arias: 2016, IBGE, Coordena¸ c˜ ao de Geografia Rio de Janeiro, Brazil, 2017

work page 2016

[25] [25]

doi:10.7910/DVN/ FAEZIO

China Data Lab, Baidu Mobility Data (2020). doi:10.7910/DVN/ FAEZIO. URL https://doi.org/10.7910/DVN/FAEZIO

work page doi:10.7910/dvn/ 2020

[26] [26]

Cota, Monitoring the number of covid-19 cases and deaths in brazil at municipal and federative units level, SciELO Preprints (2020)

W. Cota, Monitoring the number of covid-19 cases and deaths in brazil at municipal and federative units level, SciELO Preprints (2020)

work page 2020

[27] [27]

Safra Center for Ethics, Key metrics for covid suppression a framework for policy makers and the public (2020)

Harvard Global Health Institute and Harvard’s Edmond J. Safra Center for Ethics, Key metrics for covid suppression a framework for policy makers and the public (2020)

work page 2020

[28] [28]

M. Jin, H. Y. Koh, Q. Wen, D. Zambon, C. Alippi, G. I. Webb, I. King, S. Pan, A survey on graph neural networks for time series: Forecasting, classification, imputation, and anomaly detection, IEEE Transactions on Pattern Analysis and Machine Intelligence 46 (12) (2024) 10466–10485. doi:10.1109/TPAMI.2024.3443141. 27

work page doi:10.1109/tpami.2024.3443141 2024