Forecasting staple food prices in Mozambique - BSc dissertation project
# Mozambique Food Prices: Forecasting and Early-Warning Dashboard
An analytics tool that ingests, cleans, analyses and forecasts staple food
prices in Mozambique, presented through an interactive dashboard with an
early-warning signal for price spikes.
Final-year BSc Computing dissertation project (RES6011, Arden University).
## Motivation
Food-price volatility has direct consequences for food security in Mozambique,
where households spend a large share of income on food and agriculture supports
most of the population. High-quality open price data exists but is difficult
for non-specialists to interpret. This project turns that raw data into
accessible trends, regional comparisons, forecasts, and an early-warning view
of likely price spikes.
## Data sources
- **World Food Programme (VAM) Food Prices** — Mozambique staple-price records
via the Humanitarian Data Exchange (HDX).
- **World Bank Real-Time Prices (RTP)** — sub-national price and exchange-rate
data for validation and enrichment.
- **FAO GIEWS** — global price context.
All data used is secondary, aggregated, and openly licensed. The project
contains no personal or identifiable information.
## Project structure
## Methodology
The analytical work follows the CRISP-DM framework (business understanding,
data understanding, data preparation, modelling, evaluation, deployment).
The software build follows an incremental, Agile-influenced approach.
## Planned components
1. **Data pipeline** — automated ingestion, cleaning, unit and currency
normalisation.
2. **Imputation engine** — multiple methods for handling missing market-months,
evaluated by measured reconstruction error.
3. **Analysis module** — seasonal decomposition, volatility and shock
detection, regional comparison, purchasing-power metric.
4. **Forecasting module** — comparison of SARIMA, Prophet and a
gradient-boosting model with time-series backtesting.
5. **Dashboard** — interactive interface surfacing trends, forecasts, and an
early-warn …