Epidemics in High Resolution Social Networks Christopher Siu Dr. - PowerPoint PPT Presentation

Epidemics in High Resolution Social Networks Christopher Siu Dr. Theresa Migler Von-Dollen, advisor CSC 590 December 2018 1 / 8

The Copenhagen Network Study Created a “high resolution” social network of students at the Technical University of Denmark, between 2012 and 2013 Twelve million Bluetooth observations as proxy for recording face-to-face interactions Locations based on WiFi scans and GPS Frequencies of voice calls and text messages 2 / 8

Compartmental Epidemiological Models S usceptible Variations of carrier states (e.g., typhoid fever) You start out immune (e.g., MMR vs. HepB). MSIR R ecovered I nfectious E xposed (or SEIS) You can be infected but not yet infectious. SIS SEIR You’re immune after recovering (e.g., chicken pox). SIR S usceptible I nfectious S usceptible Either you’re infected or you’re not (e.g., a cold). 3 / 8

Epidemic Spreading with Mobility Patterns Traditional epidemic modeling uses reaction-difgusion equations. Granell and Mucha argue that humans do not difguse randomly through a population. Grouped population based on daytime and nighttime locations: residences or common sites. 4 / 8

Epidemic Spreading with Mobility Patterns Frías-Martínez et al. incorporated temporal location data using cell phone records from January-May of 2009 in Mexico. Modeled the 2009 H1N1 swine fmu pandemic using SEIR. Showed that government mandates restricting mobility reduced 5 / 8 virus spread by 10 % and delayed its spread by 40 hours.

Computational Complexity of Epidemic Spreading We might be interested in fjnding the most infmuential vertices. Which people need to be immunized fjrst? If we fail, which people need to be quarantined fjrst? It’s NP-hard for general models. (Surprise?) By reduction from Vertex Cover. 6 / 8

Epidemics in High Resolution Social Networks Given a high-resolution social network: Can we apply more sophisticated models to allow approximating the graph while maintaining accurate simulations? Applying: Random constructions k -core components Density decomposition … While maintaining: Rate of infection Epidemic threshold Endemic steady state … 7 / 8

Questions? https://users.csc.calpoly.edu/~cesiu/csc590/slides/tenMinEpidemics.pdf 8 / 8