Is it possible to predict global conflict? For those sired on Asimov’s work, the title of this post will be familiar. You may recall psychohistory, Isaac’s early-50s term for the mathematical mashup of history, sociology, and statistics used to make general predictions about large groups of people. To predict your personal fortunes, see either astrology or DNA analysis.
Here’s a question for you: Can the world’s largest events database could predict conflict?
Using events-tagged data from 1979 through 2012, Jay Yonamine (Penn State PhD — Data Scientist) refactored a financial forecasting machine-learning algo to digest the data, which resulted in the accurate prediction of violence across Afghaistan’s 317 districts on a monthly basis from 2008 through 2012. In my world, that’s pretty awesome!
For those so inclined to play, the current (reduced) GDELT data set is downloadable — a mere 650Mb — not suitable for an AWS micro install, but perhaps something with a tad more girth would do the trick.