Dverybest/Nigeria-Food-Insecurity-ML-Model

# Predicting Hunger Levels in Nigeria Using Food Prices Final project submission for the Kujenga program. ## Problem Statement This study develops a predictive model that uses monthly food price data from WFP market monitoring to forecast the percentage of each state's population in IPC Phase 3 or above, providing a forward-looking early warning signal ahead of official Cadre Harmonisé assessments in Nigeria. ## Research Objective To develop and evaluate an explainable machine learning framework that learns the relationship between food price fluctuations, prior assessment history, and acute food insecurity outcomes at the state level in Nigeria — using historical WFP price data and lagged Cadre Harmonisé assessments as predictors, and current Cadre Harmonisé IPC Phase 3+ classifications as the target variable — in order to enable earlier identification of states at risk of food crisis. ## Data Sources - **WFP food prices** (`data/wfp_food_prices_nga.csv` and `data/wfp_food_prices_nga (2).csv`) — World Food Programme market price monitoring data for Nigeria, covering 14 states. Two extracts are included (a sparser and a denser download); the notebook currently trains on the sparser one (`(2)`), which was found to generalize better after controlled comparison — see Methodology below. - **Cadre Harmonisé assessments** (`data/cadre_harmonise_caf_ipc_dec25.xlsx`) — official IPC Phase food-insecurity classifications for Nigeria, 2015–2025, with up to three assessment rounds per year (Jan–May, Jun–Aug, Sep–Dec). ## Repository Structure ``` ├── nigeria_food_insecurity_v2.ipynb # Notebook to run — see Methodology below ├── nigeria_food_insecurity_model.pkl # Saved final model bundle (output of Step 12) ├── data/ │ ├── wfp_food_prices_nga.csv # WFP price data (denser extract) │ ├── wfp_food_prices_nga (2).csv # WFP price data (sparser extract — used for training) │ └── cadre_harmonise_caf_ipc_dec25.xlsx # Cadre Harmo …