Reporting Scale Prediction in Conflict Event Data Using Machine Learning

Conflict event datasets play a central role in conflict forecasting and early warning systems, supporting the prediction of conflict occurrence, intensity, fatalities, and diffusion. However, comparatively little attention has been given to understanding the visibility of conflict events across reporting networks. This study investigates the feasibility of predicting the reporting scale by classifying conflict events as nationally or internationally reported using machine learning techniques. Using conflict event data from Africa, a binary classification framework was developed and evaluated across traditional machine learning, ensemble learning, and deep learning models. Ensemble-based approaches consistently achieved the strongest performance, with XGBoost obtaining an accuracy of 86.47%, an F1-score of 89.69%, and an AUC of approximately 93%. The findings suggest that the reporting scale exhibits systematic and predictable patterns that can be learned from structured conflict event data. This work extends conflict forecasting beyond the prediction of violence and severity by introducing reporting-scale prediction as a means of understanding conflict visibility and potential reporting disparities.