Transcript slides

Epidemic Alerts
EECS E6898: TOPICS – INFORMATION PROCESSING: From Data to Solutions
Alexander Loh
May 5, 2016
Epidemic Alerts
Motivation
Prior Work
Description
Experimental Setup
Requirements
Evaluation
Motivation
Infectious diseases are a problem
Early warning system could help
Actively alert users
Prior work
Google Flu Trends
BioMosaic
HealthMap
Description
Obtain geotagged event data
User submissions via the app
Hospital records
Social media posts, ie. Twitter
 Cluster data by location
 Simulate epidemic outbreaks
 Alert users
Experimental Design
Want to confirm two things
System’s predictions are accurate
System impacts public health
Informing users may change their behavior
Run system in two modes
Mode 1: does the system predict accurately?
Just gather data
Run simulation based on data
Is ground truth close to simulation?
Mode 2: does the system help users?
Alert users and gather data
Is ground truth less severe than simulation?
Requirements
App programming
A server for communication with users
Data collection/simulation may require multiple additional servers
Evaluation
Mode 1: does the system predict accurately?
Look at the geotagged data
Look at the simulated epidemic region
Do the two have a similar structure?
Mode 2: does the system help users?
Is the geotagged data roughly contained in the simulated
epidemic region?