MelvinArchie01/FEMALE-HOUSEHOLD-PREDICTION-IN-SOUTH-AFRICA

# FEMALE-HOUSEHOLD-PREDICTION-IN-SOUTH-AFRICA Female household headship has been on the rise in South Africa in recent years. Compared to male-headed households, female-headed households often face greater social and economic challenges, including higher vulnerability to poverty. The South African census collects detailed data every 10 years, but policymakers and researchers need timely and cost-effective estimates in between census years to inform interventions. This project explores how machine learning can be applied to predict the percentage of households per ward that are female-headed and living below an income threshold, using census data and other auxiliary datasets. Objectives Develop a predictive model to estimate female-headed households below an income threshold at ward level. Reduce reliance on large-scale, costly surveys between census years. Provide data-driven insights that can support policymakers in monitoring poverty and gender-related household dynamics more effectively. Methodology Data Collection & Preprocessing Census data and other socio-economic indicators were cleaned and engineered for analysis. Missing values handled, categorical variables encoded, and features scaled. Modeling Implemented boosting algorithms: LightGBM CatBoost XGBoost Models were tuned for hyperparameters and evaluated on predictive accuracy. Results CatBoost delivered the best performance in predicting female-headed households below the income threshold. Key Features Machine learning solution for population monitoring. Boosting algorithms tested and benchmarked. Policy-relevant insights into gender, poverty, and household dynamics. Scalable and adaptable to other contexts across Africa. Impact This solution demonstrates how machine learning can complement official statistics by: Reducing the cost of frequent household surveys. Improving the timeliness and accuracy of key poverty indicators. Providing evidence-based support for social policy and p …