Percentage of subscribers that move from one bts to anotherOne in the 1hour time slot
An Agent-Based Model of Epidemic Spread using Human Mobility and Social Network Information
AN AGENT-BASED MODEL OF EPIDEMICSPREAD USING HUMANMOBILITY AND SOCIAL NETWORKS E. Frias-Martinez, G. Williamson, V. Frias-Martinez Telefonica Research, Madrid, Spain email@example.com
Epidemic Disease Models Compartmental Models (SEIR)Susceptible Exposed Infectious Recovered Contact Transition Recovery Rate Rate Rate Agent Based Models Capure complexity of social interaction Limitation with the information available to generate the agents
Digital FootprintsFor the first time in human history, we have access to large-scale human behavioral data at varying levels of spatial and temporal granularities
Ce ll Phone N e t w orkCell Phone networks are built using Base Transceiver Stations (BTS)Each BTS will be characterized by a feature vector that describes t calling behavior area.
ABM for Virus Spreading using CDRMobility ModelSocial Network ModelDisease Model
M3 S3 D3 M2 S2 D2 M1 S1 D1 Social Network Mobility Model Disease Model Modelt₀ t₁ t₂ t₃ … t₉ (1 hour) Identify geographical location (BTS) Identify peers in same BTS If peer in SN then evolve disease model with p_i Else evolve disease model with p_j
Closed Reopen Preflu 27th April 6th May Alert Shutdown 17th April 1st MayMeasure the impact that government alerts had on the population Flu is very good candidate to be modelled by SEIR
Call Detail Records from 1st Jan. till 31st.May 2009 Compute mobility and social models Baseline scenario Intervention scenario Simulation April 17th to May 16th “Evolve” disease and evaluate impact in Agent’s mobility Disease transmission Spatio-temporal evolution
Agent Generation Call Detail Records from 1st Jan. till 31st.May 2009 Granularity of 1 hour 20% of slots filled /0.25 calls per hour Agents active during the different time periods Final number of agents: 25,000 Reproduction number / Latent period / infectious period obtained from the literature.
April 27th May 1st May 6th Mobility reduced between 10% and 30% Alert Closed Shutdown ReopenIntervention
Baseline (“preflu” behavior all weeks)Intervention (alert,closed,shutdown) Epidemic peak postponed 40 hours Reduced number of infected in peak agents by 10%