Data related to the quality of life and overall well-being of human populations including poverties, conflict, sanitation, and health.

human-well-being

Human Well Being

Samantha Kuzma

### Overview  
  
The Conflict Forecast dataset categorizes areas by their modeled risk of conflict over the following 12 months. Each district is labeled with one of the following: emerging conflict, ongoing conflict, or peace.    
  
 Conflict is defined by the subnational administrative district (ADM-2), forecasted to have at least 10 fatalities in the next 12 months using a random forest machine model that learns from past records and optimizes its predictive power. The forecast incorporates frequently updated environmental variables, including water indicators, in order to capture the many stressors that may lead to conflict. Updated forecasts are released on a quarterly basis.   
  
  This forecast was produced by researchers at the World Resources Institute (WRI) and the Hague Center for Strategic Studies (HCSS) under the Water, Peace, and Security Partnership (WPS).  
  
### Methodology  
  
This model forecasts conflict up to a year in advance using a random forest machine learning model. It was measured with the Armed Conflict Location & Event Data Project (ACLED) data over a 12-month period. Conflict is defined as organized violence resulting in at least 10 fatalities. Machine learning techniques are adept at detecting subtle or complex patterns within large volumes of data without requiring a priori instructions or a theoretical framework. Since conflict literature does not offer a clear consensus on what factors contribute to violence, nor on the causal pathways to conflict, machine learning was identified as a viable means of establishing quantitative linkages between local conditions and the subsequent emergence (or continuation, or cessation) of conflict. 
  
  

The data used as indicators in the model capture a wide range of factors—from socioeconomics to governance to biophysical phenomena—and are correspondingly collected at a variety of geographic and temporal resolutions.
In order to utilize and analyze them in concert, data was reconciled to a common unit of analysis: the ADM-2 month.  Only districts with at least three years of recorded organized violence within the selected conflict database were eligible for inclusion in the modeling process.  Currently, the geographic scope is limited to the following regions: Africa, Asia, and the Middle East.
  
  

The model used historical data to learn to “predict” past conflict outcomes based on past conditions. The model was then applied to a separate tranche of more recent historical data and used its performance on this validation data to estimate the model’s predictive power on “true” forecasts made based on current conditions.  When applied to test data, the model captures 86 percent of future conflicts. The model’s conflict signal is noisy, with half of conflict predictions representing false positives.  
  
### Additional Information  
  
Resource Watch shows only a subset of the dataset. For access to the full dataset and additional information, click on the “Learn more” button.  
  
### Visualizing the Data  
  
Our team reformatted this dataset before displaying it on Resource Watch. See the documentation on how Resource Watch retrieved the data on [Github](https://github.com/resource-watch/data-pre-processing/tree/master/wat_066_rw0_conflict_forecast). This data has been mapped on Resource Watch by using a shapefile provided by the data source.  
  
### Disclaimer  
  
Excerpts of this description page were taken from the source metadata.

Hague Center for Strategic Studies (HCSS)

Water, Peace, and Security Partnership (WPS)

Water, Peace and Security Global Early Warning Tool

wat-066-rw0-conflict-forecast-rw

Long-term Conflict Forecast: August 2025 - July 2026

A forecast for peace or risk of emerging or ongoing conflict with at least 10 fatalities over the next 12 months

No Team

Conflict Forecast

Long-term Conflict Forecast: August 2025 - July 2026