IDAS Seminars

20.3.2020 Data Science Colloquium: Conflict Modelling with Multi-Layer Networked Dynamics: Uncovering Interaction Structure and Causal Factors

by Weisi Guo (Turing Fellow, Cranfield University)

218 (OC)




Conflict plagues human development and the nature of armed violence has transformed since the end of the Cold War. Today, political violence is trans-national, interleaved with criminal enterprises, and highly stochastic. Traditional conflict prediction approaches using human knowledge, statistical laws, and agent-based models can no longer scale to match an evolving reality across complex landscapes. The Alan Turing Institute has been running a 2 year program (GUARD) to create fundamentally new mathematical frameworks to describe conflict dynamics and analyse the topological attributes of our connected world. We have developed a new global conflict model at city-scale resolution that demonstrates powerful in-sample accuracy and out-of-sample predictive power across diverse geographic regions and genres. It paves the way towards explainable power and we are now integrating this framework with advanced AI and climate change models to inform the UK government.