Researchers have developed a machine learning approach to audit the selection process for domestic violence surveys within the Demographic and Health Surveys (DHS) program across South Asia. By analyzing demographic, socioeconomic, and household characteristics from over 796,000 women in Afghanistan, India, Myanmar, Nepal, and Pakistan, they identified six distinct profiles. The HistGradientBoosting model achieved a high ROC-AUC of 0.9167, significantly outperforming a baseline model and demonstrating that interpretable machine learning can enhance the auditing of sensitive survey modules beyond traditional sampling designs. AI
IMPACT Demonstrates how machine learning can improve the rigor and transparency of sensitive data collection processes.
RANK_REASON Academic paper detailing a novel application of machine learning for auditing survey data. [lever_c_demoted from research: ic=1 ai=1.0]
- Afghanistan
- Demographic and Health Surveys
- HistGradientBoosting
- India
- Mohammad Abu Obaida Mullick
- Myanmar
- Nepal
- Pakistan
- South Asia
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →