Ce diaporama a bien été signalé.
Nous utilisons votre profil LinkedIn et vos données d’activité pour vous proposer des publicités personnalisées et pertinentes. Vous pouvez changer vos préférences de publicités à tout moment.
Optimizing

sentinel

surveillance

in

static

and

temporal

networks
Petter

Holme
Tokyo

Institute

of

Technology
P. Holme, Objective measures for
sentinel surveillance in network
epidemiology, arXiv:1803.10637
Medical epi background
Active surveillance
Contact tracing
Passive surveillance
Hospitals/units reporting all stats
Sentinel surveillance
Epidemi...
Sentinel surveillance
Modeling background
Vaccination
Nodes that, if deleted, minimize
outbreaks
Sentinel surveillance
Nodes that get infected reliably and
early
In...
Time to discovery or extinction
Frequency of detection
Time to discovery
. . given that it is discovered
Objective measures
Goal
Compare these thee objective
measures.
How
• Run the (standard, continuous time,
Markovian) SIR model on empirical
te...
time; t is the time resolution of the data set, M is the number of links in the projected and thresholded static networks,...
Results
Time to detection or extinction Time to detection Frequency of detection
Nodesoftopimportance:
3
2
1
Objective measures
fo...
Correlations between objective
measures, static networks
St Andrews
Dating
E-mail 3
Reality
Romania
High School 5
Gallery ...
Ditto,temporalnetworks
Primary School 1
Diary
Gallery 3
High School 3
E-mail 3
Conference 2
Cambridge 2
Dating Copenhagen ...
Structural correlations
Static networks
Subgraph centrality Component size Harmonic closeness Harmonic vitality Coreness
G...
Subgraph centrality
Component size
Harmonic closeness
Harmonic vitality
Coreness
Degree
Timetodetection
orextinctionTimeto...
Primary School 1
Diary
Gallery 3
High School 3
E-mail 3
Conference 2
Cambridge 2
Dating Copenhagen Calls
Romania
Gallery 1...
Conclusions
•Choose objective measure depending on
what aspect of sentinel surveillance that
matters (because of extinctio...
Prochain SlideShare
Chargement dans…5
×

Optimizing
 sentinel
 surveillance 
in
 static
 and 
temporal 
networks

606 vues

Publié le

talk at the Contagion & Networks Satellite Sympoisum, NetSci 2018

Publié dans : Sciences
  • Hello! I can recommend a site that has helped me. It's called ⇒ www.HelpWriting.net ⇐ So make sure to check it out!
       Répondre 
    Voulez-vous vraiment ?  Oui  Non
    Votre message apparaîtra ici
  • Hello! I have searched hard to find a reliable and best research paper writing service and finally i got a good option for my needs as ⇒ www.WritePaper.info ⇐
       Répondre 
    Voulez-vous vraiment ?  Oui  Non
    Votre message apparaîtra ici
  • Get Now to Read eBook === http://readjpaojdpa.ygto.com/3642364608-temporal-networks.html
       Répondre 
    Voulez-vous vraiment ?  Oui  Non
    Votre message apparaîtra ici
  • Soyez le premier à aimer ceci

Optimizing
 sentinel
 surveillance 
in
 static
 and 
temporal 
networks

  1. 1. Optimizing
 sentinel
 surveillance
 in
 static
 and
 temporal
 networks Petter
 Holme Tokyo
 Institute
 of
 Technology
  2. 2. P. Holme, Objective measures for sentinel surveillance in network epidemiology, arXiv:1803.10637
  3. 3. Medical epi background
  4. 4. Active surveillance Contact tracing Passive surveillance Hospitals/units reporting all stats Sentinel surveillance Epidemiology is the study of the distribution and determinants of health & disease conditions in defined populations.
  5. 5. Sentinel surveillance
  6. 6. Modeling background
  7. 7. Vaccination Nodes that, if deleted, minimize outbreaks Sentinel surveillance Nodes that get infected reliably and early Influence maximization Nodes that maximize outbreaks. Interesting before an outbreak of a known pathogen happens Makes no sense for n = 1 Importance concepts
  8. 8. Time to discovery or extinction Frequency of detection Time to discovery . . given that it is discovered Objective measures
  9. 9. Goal Compare these thee objective measures. How • Run the (standard, continuous time, Markovian) SIR model on empirical temporal and static networks of human contacts. • Compare the outcome to each other and explain differences by network structure. Outline of this research
  10. 10. time; t is the time resolution of the data set, M is the number of links in the projected and thresholded static networks, and ✓ is the threshold. Data set N C T t M ✓ Ref. Conference 1 113 20,818 2.50d 20s 1,321 2 [16] Conference 2 198 327,333 2.95d 20s 775 75 [17] Hospital 75 32,424 4.02d 20s 582 8 [18] Reality 63 26,260 8.63h 5s 421 3 [19] O ce 92 9,827 11.4d 20s 389 3 [20] Primary School 1 236 60,623 8.64h 20s 299 3 [21] Primary School 2 238 65,150 8.58h 20s 257 3 [21] Romania 42 1,748,401 62.8d 1m 128 61 [22] High School 1 312 28,780 4.99h 20s 1,385 2 [23] High School 2 310 47,338 8.99h 20s 1,601 2 [23] High School 3 303 40,174 8.99h 20s 1,096 3 [23] High School 4 295 37,279 8.99h 20s 1,363 2 [23] High School 5 299 34,937 8.99h 20s 1,298 2 [23] Gallery 1 200 5,943 7.80h 20s 398 2 [24] Gallery 2 204 6,709 8.05h 20s 393 2 [24] Gallery 3 186 5,691 7.39h 20s 362 2 [24] Gallery 4 211 7,409 8.01h 20s 294 2 [24] Gallery 5 215 7,634 5.61h 20s 967 1 [24] Cambridge 1 186 3,853,714 6.07d 20s 180 1,312 [17] Cambridge 2 2,536 2,064,114 3.89d 1s 5,996 42 [25] Intel 112 2,448,720 4.15d 20s 107 1,326 [17] Copenhagen Bluetooth 671 458,920 28.0d 20s 13,363 2 [26] Kenya 52 2,070 2.54d 1h 43 26 [27] Diary 49 2,143 418d 1d 345 4 [28] Prostitution 16,730 50,632 6.00y 1d 39,044 1 [29] St Andrews 25 408,996 74d 1s 139 379 [30] WiFi 18,719 9,094,619 83.7d 5m 884,800 6 [31] Facebook 45,813 855,542 1,561d 1s 183,412 1 [32] Messages 35,624 489,653 3,018d 1s 94,768 2 [33] Forum 7,084 1,429,573 3,141d 1s 70,942 2 [33] Dating 29,341 529,890 512d 1s 74,561 2 [34] Copenhagen Calls 483 10,545 28.0d 1s 271 6 [26] Copenhagen SMS 533 30,380 21.6d 1s 320 12 [26] E-mail 1 57,194 444,160 112d 1s 92,442 1 [35] E-mail 2 3,188 309,125 81d 1s 16,220 3 [36] E-mail 3 986 332,334 526d 1s 9,474 3 [37] E-mail 4 167 82,927 271d 1s 1,830 4 [38] College 1,899 59,835 193d 1s 8,608 2 [39] Datasets
  11. 11. Results
  12. 12. Time to detection or extinction Time to detection Frequency of detection Nodesoftopimportance: 3 2 1 Objective measures for the Office data at ß = 1
  13. 13. Correlations between objective measures, static networks St Andrews Dating E-mail 3 Reality Romania High School 5 Gallery 5 Kenya Facebook Copenhagen SMS College High School 1 Gallery 1 Cambridge 1 Diary Messages E-mail 1 Conference 1 Office High School 2 Gallery 2 Cambridge 2 Prostitution Forum E-mail 2 Conference 2 Primary School 1 High School 3 Gallery 3 Intel Hospital Primary School 2 High School 4 Gallery 4 Copenhagen Bluetooth WiFi Copenhagen Calls E-mail 4 1010-3 10-2 0.11 1 –1 0 timetodiscoveryvstimetodiscoveryordetection timetodiscoveryvsfrequencyofdiscovery ß T
  14. 14. Ditto,temporalnetworks Primary School 1 Diary Gallery 3 High School 3 E-mail 3 Conference 2 Cambridge 2 Dating Copenhagen Calls Romania Gallery 1 High School 1 E-mail 1 Reality Messages Copenhagen SMS Office Kenya Gallery 2 High School 2 E-mail 2 Conference 1 Cambridge 1 Forum Primary School 2 Prostitution Gallery 4 High School 4 E-mail 4 Hospital Intel College St Andrews Gallery 5 High School 5 Copenhagen Bluetooth FacebookWiFi 0.01 10.1 0.01 0.1 1 vT –1 –0.5 0 0.5 1 ß T
  15. 15. Structural correlations Static networks Subgraph centrality Component size Harmonic closeness Harmonic vitality Coreness Gallery 5 High School 5 Reality Copenhagen SMS Facebook Romania Kenya Gallery 2 High School 2 Conference 1 Prostitution Forum Office Cambridge 2 E-mail 2 Gallery 3 High School 3 Conference 2 St Andrews Dating Primary School 1 Intel E-mail 3 Gallery 4 High School 4 Hospital Copenhagen Calls College Primary School 2 Copenhagen Bluetooth E-mail 4 Gallery 1 High School 1 Diary Messages Cambridge 1 E-mail 1 Degree WiFi 0.10.01 10.001 10 Time to detection or extinction Time to detection Frequency of detection ß
  16. 16. Subgraph centrality Component size Harmonic closeness Harmonic vitality Coreness Degree Timetodetection orextinctionTimetodetectionFrequencyofdetection
  17. 17. Primary School 1 Diary Gallery 3 High School 3 E-mail 3 Conference 2 Cambridge 2 Dating Copenhagen Calls Romania Gallery 1 High School 1 E-mail 1 Reality Messages Copenhagen SMS Office Kenya Gallery 2 High School 2 E-mail 2 Conference 1 Cambridge 1 Forum Primary School 2 Prostitution Gallery 4 High School 4 E-mail 4 Hospital Intel College St Andrews Gallery 5 High School 5 Copenhagen Bluetooth FacebookWiFi Downstream component size Average time First time Degree Duration Strength Upstream component size 0.01 10.1 0.01 0.1 1 vT ß
  18. 18. Conclusions •Choose objective measure depending on what aspect of sentinel surveillance that matters (because of extinction events that can make a big difference). Thank you!

×