notes.txt

1st meeting with nils - 
3 papers to review - 1st step is to investgate unit of scale
- either network, cluster, or geogrid approach - can talk to sanjeev 
about this 

04/06/2019
TODO: 
organise weekly meeting with sanjeev - send some overview papers across.
get on same page 
organise meet up with nils and sanjeev 
- look into data sets - acled icews prio ucdp. 
- need to learn sql? 
EMAIL NILS ABOUT DROPBOX DONE

VIEWS: A POLITICAL VIOLENCE EARLY-WARNING SYSTEM.
V GOOD AT PREDICTION AT LOWER LEVEL - PGM 99.1% ACCURACY 
LOWER ON CM GRID LEVEL PREDICTION IS 93.1 OR WHATEVER.
SAYS NOT AS GOOD AT PREDICTING START OF CONFLICT - POTENTIAL IN???
ONLY IN AFRICA. 
different ensembling methods.


05/06/2019
TODO
LOOK AT MATPLOTLIB PLOTTING ECT. 
send nils mark scheme and requirements for the project.

CLONE VIEWS BRANCH - HAVE A PLAY WITH THE DATA.
osa - one step ahead has most of the iunteresting stuff in it. 

SPATIAL ECOLOGY OF WAR AND PEACE
uses betweenness / degree as a predictor (strategic centrality). use this to predict 74% of conflict from centrality > 10^4
uses agent based models in culture swapping game to find the culture clusters - some flicker back and forth between dominant cultures - these are then ones in whihc have high strategic centrality and ahve a high risk of terorism.

data is from - global terrorism database (GTD)
Prio - ucdp database for non terrorism.


06/06/2019
Look at the data
icews - narrow down which are being used 
alced 
prio - only for africa?
ucdp
david cunningham maryland, zhukov epidemic - final - interesting data manipulation. 
decide what type of machine learning to apply - making pipelines ect. 
GOT CARTOPY INSTALLED. 
LOOK AT NOE PAPER - WRITE ONE PAGE OF LIT REVIEW.