Forecasting Internally Displaced Population Migration Patterns in Syria and Yemen

Area of Science:

• Disaster response
• Computational social science
• Humanitarian logistics

Background:

• Armed conflict drives unprecedented internal displacement, creating urgent needs for shelter, food, and healthcare.
• Predicting internally displaced persons (IDPs) migration patterns is crucial for effective aid delivery but remains a significant challenge.
• Humanitarian organizations require robust forecasting tools to allocate resources efficiently during crises.

Purpose of the Study:

• To develop and evaluate a machine learning framework for forecasting internally displaced persons (IDPs) migration.
• To empower humanitarian aid groups with predictive capabilities for proactive resource allocation.
• To improve the timeliness and effectiveness of aid delivery to vulnerable populations.

Main Methods:

• Modeled monthly IDP migration within Syria and Yemen using diverse datasets.
• Included variables such as food prices, fuel prices, wages, location, time, and conflict reports.
• Compared machine learning forecasting models against baseline persistence methods.

Main Results:

• A machine learning approach, specifically a random forest model, demonstrated superior accuracy in forecasting IDP migration trends.
• The random forest model outperformed the best persistence model by 26% for Syria and 17% for Yemen in terms of root mean square error of log migration.
• This indicates a significant improvement in predictive performance over traditional forecasting techniques.

Conclusions:

• Integrating multiple data sources into machine learning models enhances the prediction of internally displaced persons (IDPs) migration.
• Further research is needed to assess the practical implementation of these models for proactive aid allocation.
• Successful implementation could enable humanitarian efforts to anticipate and respond to forecast population arrivals more effectively.