Eliya-Nandi/Help-International-Aid-Prioritization

Unsupervised machine learning project using PCA, K-means and hierarchical clustering to prioritize countries for Humanitarian Aid Allocation. # HELP International: Aid Prioritization Using Unsupervised Learning ## Project Overview HELP International, a humanitarian NGO, raised **$10 million** and needed a data-driven way to identify countries that should receive aid first. This project applies **unsupervised machine learning** to group countries based on socio-economic and health indicators, then recommends the countries with the highest need for funding support. The analysis combines **data preprocessing, exploratory data analysis, PCA, K-means clustering, hierarchical clustering, silhouette validation, cluster profiling, and need-score ranking**. --- ## Live Report View the knitted HTML report here: **Open the Professional HTML Report** --- ## Business Problem Aid allocation is a multidimensional decision. A country may require support because of high child mortality, low GDP per capita, weak health expenditure, low income, poor life expectancy, or limited economic capacity. The goal of this project is to help HELP International answer: > **Which countries should be prioritized for aid based on socio-economic and health indicators?** --- ## My Role **Data Scientist / Machine Learning Analyst** I handled the complete analytical workflow: - Cleaned and prepared the country-level socio-economic dataset - Engineered policy-relevant variables by converting trade and health percentages into GDP-based actual values - Performed exploratory data analysis to understand skewness, scale differences, and outliers - Applied PCA for dimensionality reduction - Compared hierarchical clustering and K-means clustering - Evaluated cluster quality using silhouette scores - Profiled clusters using GDP, income, health expenditure, exports, and child mortality - Built a need-score ranking to identify the highest-priority countries for aid allocation - Translated model outputs into actionable recommendations for decision-makers --- ## Dataset The dataset contains **167 countries** and **9 numeric socio-econ …