Biomedical Engineering Reference
In-Depth Information
information of the profile, which typically includes, among other things, name, gen-
der, age, affiliations (school, work, region), birthday, location, education history, and
friends.
The location field helps us in tracking the current/default location of a user. Geo
location codes are present in a location enabled mobile tweet/post. For all other pur-
poses, we assume the location attribute within the profile page to be his/her current
location and pass it as an input to Google's location based web services to fetch geo-
location codes (i.e., latitude and longitude) along with the country, state, city with a
certain accuracy scale. All the data extracted from posts and profile page are stored in a
spatio-temporal “OSN data” Database.
We apply filters to get quantitative data within Unites States and exclude organiza-
tions and users who posts multiple times during a certain period of time on flu related
activities. This data is fed into the Analysis Engine which has a detector and ARX pre-
dictor model. The visualization tools and reporting services generate timely visual and
data centric reports on the ILI situation. The CDC monitors Influenza-like illness cases
within USA by collecting data about number of Hospitalizations, percentages weighted
ILI visits to physicians, etc, and publishes it online. We download the CDC data into
“ILI data” database to compare with our results.
4DaaSt
In this section we briefly describe our datasets used for influenza prediction. OSN has
emerged as a primary source of user interactions on daily events, health status up-
dates, entertainment, etc. At any given time, tens of millions of users are logged onto
OSN's, with each user spending an average of tens of minutes daily. Since Oct 18,
2009, we have searched and collected tweets and profile details of Twitter users who
mentioned flu descriptors in their tweets. Facebook opened their Search functionality
in early February 2010 and since then we have been fetching status updates and wall
posts of Facebook users with mention of flu descriptors. The preliminary Twitter anal-
ysis for the year 2009-2010 is documented in [1]. For 2010-2011, we have 4.5 million
tweets from 1.9 million unique users and 2.0 million facebook posts from 1.5 million
unique facebook users. Twitter allows its users to set their location details to public or
private from the profile page or mobile client. So far our analysis on location details
of the Twitter dataset suggest that 22% users on Twitter are within USA, 46% users
are outside USA and 32% users have not published their location details. Analysis on
location details of Facebook dataset suggest that 22% users are within USA, 17% users
are outside USA and 61% users have not published their location details.
Initial analysis for the period 2009-2010 indicated a strong correlation between CDC
and Twitter data on the flu incidences [1]. However results for the year 2010-2011
showed a significant drop in the correlation coefficient from 0.98 to 0.47. In an attempt
to investigate such a drastic drop in correlation we looked at data samples and found
spurious messages which suppressed the actual data. To list a few, tweets like “I got flu
shot today.”, “#nowplaying Vado - Slime Flu..i got one recently!” (Slime flu is the name
of a debut mixtape from an artist V.A.D.O. released in 2010) are false alarms. In the year
2009-2010, the Swine Flu event was so evident that the noise did not significantly affect
 
Search WWH ::




Custom Search