Skip to content

Latest commit

 

History

History
43 lines (33 loc) · 2.78 KB

README.md

File metadata and controls

43 lines (33 loc) · 2.78 KB

WISDOM

(Widespread Intelligent System for Domain Outlier Monitoring)

WISDOM uses scipy's implementation of Kernel Density Estimation (KDE) to build density estimates for daily counts of weapon ads faceted by weapon type and city (both extracted using NER from crawled weapons data). Daily counts that have low densities are flagged as anomalous and indicated in the interface by a pulsing red circle. Density cutoff for an anomaly can be adjusted by user in interface.

alt tag

By hovering over circles, density estimates, histograms, and trends for given location/weapon combos can be visualized in a d3 tooltip. The most recent day's count (used for anomaly evaluation) and anomaly status is indicated by a vertical line through the desnity estimate.

alt tag

Notes on data

  • For this prototype, ads found on days between 9-20-15 and 10-19-15 were queried as this was the primary period when data was crawled by the Memex Weapons team.
  • Not all weapon type / city combos have been included
  • Kernel densities have been manually inflated due to limited data (less data -> less certainty about distribution -> low densities)

Setup

To-Dos and Ideas

  • Allow user to set time window (if more crawl data becomes available)
  • Allow analysis by country as well as by city
  • U.S.-only interface
  • Make eastern seaboard cities less congested in viz
  • Scale weapons buttons to hundreds of selections
  • Evaluate performance around known weapon events