KODAK-CODER/child_mortality_in_kenya

ML recommendation system for Kenya child mortality # Child Mortality Recommendation System — Kenya > A machine learning system that classifies Kenya's 47 counties by child mortality risk tier and generates ranked, evidence-based public health intervention recommendations tailored to each county's unique profile. --- ## Table of Contents - Project Overview - Business Problem - Dataset - Project Structure - Getting Started - How to Run - Models - Results - Visualisations - Repository Structure - Technologies Used - Author** masks a **3.5× disparity** between best and worst performing counties — ranging from 22.1 in Nairobi to 78.2 in Mandera. With limited health budgets, county health teams need data-driven guidance on which interventions will have the greatest impact in their specific context. This project develops a **hybrid recommendation system** that: 1. **Classifies** all 47 Kenyan counties into High, Medium, or Low mortality risk tiers 2. **Predicts** county-level under-5 mortality rates from socioeconomic and health indicators 3. **Recommends** the top-3 most effective, feasible, and context-appropriate public health interventions for each county The system is aligned with Kenya's MOH RMNCH strategy and the **UN Sustainable Development Goal 3.2** (reduce under-5 mortality to ≤25 per 1,000 by 2030). --- ## Business Problem County Health Management Teams across Kenya face two key challenges: - **Information overload** — dozens of possible interventions with varying evidence levels - **Resource constraints** — limited budgets require precise targeting of the highest-impact actions This system solves both by providing **personalised, ranked, county-specific intervention recommendations** grounded in both peer-county learning (collaborative filtering) and evidence-based intervention scoring (content-based filtering). **Stakeholders:** MOH Kenya RMNH Division · 47 County Health Management Teams · WHO UNICEF --- ## Dataset Three datasets are used in this project: | Dataset | Source | Records | Descri …