Predicting Household Vulnerability to Food Insecurity in East Africa Using Socioeconomic and Climate Data
# 📊 Predicting County-Level Acute Food Insecurity in Kenya
### A Climate–Market–Conflict Early Warning Model
---
## 🌍 Project Overview
This project develops a **county-level early warning system** to predict acute food insecurity (IPC Phase 3 or worse) in Kenya one month ahead. **streamlit Dashboard.**
huggingface.co
Current food security assessments under the :contentReference[oaicite:0]{index=0} (IPC) primarily describe past conditions. This limits proactive humanitarian action.
By integrating:
- Climate indicators
- Market prices
- Conflict data
this project builds a predictive model that supports forward-looking humanitarian decision-making.
The analytical framework aligns with institutions such as:
- World Food Programme(WFP)
- Food and Agriculture Organization(FAO)
- Famine Early Warning Systems Network(FEWS NET)
---
## 🎯 Objectives
### Primary Objective
Build a county-level model that estimates the probability of entering **IPC Phase 3+** one month ahead.
### Secondary Objectives
- Construct a harmonized county–month panel dataset
- Identify the most important drivers of acute food insecurity
- Compare:
- Logistic Regression
- Random Forest
- XGBoost
- Develop a prototype dashboard for risk visualization
---
## 🧠 Problem Statement
Food insecurity monitoring in Kenya is largely descriptive and retrospective. There is no simple, integrated predictive system combining:
- Rainfall variability
- Vegetation health
- Food prices
- Conflict instability
As a result:
- Humanitarian response is reactive
- Resources are allocated after crisis onset
- Early warning signals are not fully utilized
This project addresses that gap with a **short-term classification model for IPC Phase 3+ risk.**
---
## 📦 Data Sources
All datasets were harmonized into a **county–month panel (2015–2025).**
### 1️⃣ IPC Classifications (Target Variable)
Source:
ipcinfo.org
- Original quarterly IPC phases expande …