In a systematic review of civil-war prediction research, every paper claiming that complex machine-learning models outperformed older statistical models failed to reproduce because of data leakage.
Notes on verification
Confirmed by peer-reviewed Patterns article (Kapoor & Narayanan 2023) and consistent preprint/PMC versions detailing systematic review of civil-war prediction ML papers and their data leakage issues.
Sources
- Leakage and the reproducibility crisis in machine-learning-based science (cell.com)
- https://pmc.ncbi.nlm.nih.gov/articles/PMC10499856/ (pmc.ncbi.nlm.nih.gov)
- https://pubmed.ncbi.nlm.nih.gov/37720327/ (pubmed.ncbi.nlm.nih.gov)
- https://arxiv.org/pdf/2207.07048 (arxiv.org)