Skip to content

Latest commit

 

History

History
43 lines (20 loc) · 1.38 KB

OLD.md

File metadata and controls

43 lines (20 loc) · 1.38 KB
Raw

Twitter Movement Derivation

Given a Twitterer's geotagged Twitter history, can we:

  1. Determine a home location?
  2. Determine if they are geographically at-risk during a hazardous weather event?

Method

1. Data Acquisition

Data is retrieved from GNIP for the East Coast during Hurricane Sandy. This process is further outlined in the Data_Aquisition Directory. All of these tweets may be viewed here: epic.cs.colorado.edu/Twitter-Movement-Derivation/dataset

2. GeoProcessing

3. TimeProcessing

  1. Data is then cleaned and processed with Spark to identify users with (1) tweet in ZoneA. And then clustered with DBScan. The details of this are outlined in the GeoProcessing Directory.

  2. TimeProcessing then looks at every user's temporal spread over a given week and validates the spatial clustering patterns.

  3. The TileProcessing directory filters all of the tweets by time (before, during, and after), and creates tilesets for visualizations.

  4. Shelter-In-Place looks for users who stayed.

  5. Evacuation looks for users who left at some point before landfall.

ScratchPad

Total Tweets (from all Jobs): 3,658,714 This number should be the sum of all the individual jobs, will have to confirm with the Dropbox file

Total Tweets with Geo Tag: 3,632,625... wtf?