Conflict Forecast: a new tool to predict conflict before it happens

Barcelona School of Economics (BSE)
  • A tool for the prediction of risk of violence and armed conflict, Conflict Forecast, is developed by Barcelona GSE and University of Cambridge researchers
  • Supervised and unsupervised machine learning approaches on a database of 5 million newspapers enable Conflict Forecast to predict conflict even in nations with little or no track record of instability
  • Conflict Forecast may prove to be an invaluable tool in decision making and international deployment of policies to prevent conflict and its destructive effects


In international news outlets, particularly those addressing matters of geopolitics, news regarding political instability and social unrest make common headlines. It is no wonder that this is so, as events of political and social turmoil or full-blown conflict undermine not only the stability of countries, but put their social fabric under severe stress. In turn, the influence on neighboring countries is another factor not to be overlooked.

In connection with this, researchers Hannes Mueller (IAE-CSIC and Barcelona GSE) and Christopher Rauh (University of Cambridge) have launched a website providing forecasts for outbreaks of violence and escalation into armed conflict. Being able to predict the risk for such events has obvious benefits, notably, to provide a framework for preventive measure decision making. In addition, early warning may prove very valuable to take action in order to prevent situations escalating out of control.


Conflict Forecast: the tool

The predictions of Confict Forecast are based on a database of more than 5 million newspaper articles which are updated continuously. The research team developed their forecast method over the course of more than five years to extract subtle signals from the news, even in low-risk countries without a recent history of conflict.

Street violence, in addition to its negative effects, can have the potential to escalate into a full-blown armed conflict.

The website makes the team's method accessible to a broader public through cutting-edge data visualization tools. It also gathers forecast histories, which are available to practitioners and research teams worldwide. Forecasts are updated regularly and can be downloaded easily from the Conflict Forecast website itself.

The technical details of the framework were published at Reading Between the Lines: Prediction of Political Violence Using Newspaper Text in American Political Science Review, also at the Barcelona GSE Working Paper The Hard Problem of Prediction for Conflict Prevention, with a lay summary available at Barcelona GSE Focus: Can machine learning help policymakers detect conflict?

A technical hurdle overcome by the researchers in this work, which is very much worth highlighting, is the forecast of conflict in countries with a long record of absence of conflict. The low baseline risk of stable countries with little or no history of serious conflict influences the expectations and potential predictions which a model can provide. This problem was overcome by using a large volume of newspaper information via supervised and unsupervised machine learning.

A snapshot of the Conflict Forecast website. One can find further details on the forecast by clicking on each country.

The authors note that forecasts do not provide a causal analysis of the underlying factors of high risk but only produce a warning of that risk. Additional analysis of the specific circumstances is needed to identify ways to address the conflict risk.

The output forecast is to be evaluated to decide on the optimal intervention that needs to be taken to minimize the total cost of conflict plus intervention. This cost-benefit evaluation exercise highlights the potential cost savings of prevention, for which reliable forecasts are a prerequisite.

Being able to prevent severe hazard before it materializes has the potential to yield enormous benefits. Conflict Forecast can provide international decision-makers and country experts with an objective benchmark to mobilize de-escalation and peace processes well in time.


Image credits:

Street riots picture was downloaded from Flickr and licensed via a Creative Commons Attribution 2.0 Generic (CC BY 2.0) license.

Conflict Forecast snapshot reproduced under fair use terms, taken directly at the Conflict Forecast website.

Related articles