Predictive Model For Humanitarian Aid - Research on a conflict early warning system for the Sahel region

This study explores the complex situation in the Sahel region, an area characterized by continuous social conflicts and significant political instability. It presents an early warning system developed to predict violence across 12 Sahelian countries, focusing on the urgent need for accurate forecasting to guide policy decisions and predict catastrophic events. Utilizing historical data, the system predicts fatalities at both national and regional level, in order to enable policymakers and stakeholders to take appropriate action in a timely manner, given the expected conflict. Central to the study is the integration of data from multiple sources. In particular, GDELT provides structured information extracted from online news articles relevant to regional conflict dynamics, while ACLED offers a comprehensive dataset on political violence, protests, and associated events worldwide. Synthetic data is further generated through state-of-the-art causal data augmentation methods to enable the model to be more adaptive to changes in scenarios, including wartime and peacetime contexts. Model calibration is one of the critical elements in this predictive framework, ensuring that confidence levels associated with the forecasts actually represent the true likelihood of future events. Through the application of calibration techniques, such as conformal prediction, the model accounts for the unique temporal dynamics and potential covariate shifts inherent in the data, improving the reliability of predictions. Various temporal granularities (daily, weekly, and monthly) are explored in forecasts, enhancing the system’s responsiveness to both immediate crises and long-term trends. Another consideration is the ethical responsibility to deploy predictive models related to conflict-prone countries. This research emphasizes data minimization whenever possible and encourages reflection on the broader societal implications of predictive analytics. Addressing those technical and ethical dimensions, this study aims to contribute to an effective and humane approach in managing and mitigating the conflicts in the Sahel region.