BoldwinMax/Zambia-FRA-Maize-Price-Prediction
Domaine:
agriculturesocioeconomicType de record:
project
Créateur:
Bol
Hôte:
Predicting Zambia's FRA maize floor price ahead of the annual announcement, using live World Bank & FRED data. Interactive Streamlit dashboard included.
# 🌽 Zambia FRA Maize Floor Price - Predicting the 2026 Announcement
Predicting the **Food Reserve Agency (FRA)** maize floor price (Kwacha per 50 kg
bag) ahead of its annual announcement, using Zambian and global economic data
from 2002–2025.
**Headline result:** the model points to a 2026 floor price in the **mid-K350s**
(indicative range ~K341–K371) - a modest, single-digit-percentage rise on the
2025 price of K340.
> ⚠️ A small-sample, indicative model - see the caveats below. The FRA price is
> ultimately a policy decision.
## 🔗 Live dashboard
**➡️ Interactive dashboard**
Move the sliders for the 2026 exchange rate, harvest, and inflation, and watch
the predicted price update live.
## 📂 What's in this repo
| File | What it is |
|------|------------|
| `Zambia-FRA-Maize-Price-Prediction.ipynb` | The full analysis — EDA, modelling, and verdict (renders directly on GitHub) |
| `dashboard.py` | Interactive Streamlit dashboard (the "what-if" tool) |
| `requirements.txt` | Python dependencies |
## 📊 Data sources
| Feature | Source | Access |
|---------|--------|--------|
| Inflation (annual %) | World Bank API (`FP.CPI.TOTL.ZG`) | live |
| Exchange rate (ZMW/US$) | World Bank API (`PA.NUS.FCRF`) | live |
| Global maize price ($/t) | FRED API (`PMAIZMTUSDA`) | live |
| FRA floor price | FRA announcements / academic records | embedded |
| Maize harvest | ZamStats Crop Forecast Survey | embedded |
| Fertiliser (DAP) | World Bank Pink Sheet | embedded |
Live data is fetched at runtime, with embedded values as a fallback so the
project always runs offline. All prices are in today's rebased kwacha (pre-2013
figures divided by 1,000).
## 🔬 Method
A **Ridge regression** (regularised linear regression) on five features -
previous year's price, exchange rate, inflation, maize harvest, and an
election-year flag - evaluated with **leave-one-out cross-validation** (≈24
annual observations). The previous year's price and the exchange rate dominate;
the FRA effectively …