Mi-kami/Early-Warning-Food-Price-Monitor-Northeast-Nigeria

A capstone project marking the completion of the TS Academy Data Science Programme — showcasing four months of rigorous learning, real-world problem solving, and our readiness to take our first official steps into the data science field. # 🌍 Early Warning Food Price Monitor — Northeast Nigeria > A multi-model time series forecasting system for predicting monthly retail food prices up to three months ahead across conflict-affected states in Northeast Nigeria. --- ## 🔗 Live Deployment **👉 earlywarningfoodpricemonito… The app provides a three-month price outlook for five key commodities across Adamawa, Borno and Yobe — three conflict-affected states in Northeast Nigeria where food insecurity is most acute. No technical knowledge required. --- ## 📌 Project Overview Food price volatility in Northeast Nigeria is driven by a compounding set of structural forces: the Boko Haram insurgency, the June 2023 fuel subsidy removal, naira devaluation, seasonal harvest cycles, and climate variability. Existing price monitoring systems report prices retrospectively — after market changes have already occurred. This project addresses that gap by building a **data-driven forecasting pipeline** that predicts monthly retail food prices **up to three months in advance**, giving policymakers, humanitarian organisations, traders and farmers the lead time needed to act before a crisis arrives. **Three models were built, evaluated and compared:** - SARIMA — classical seasonal time series model - XGBoost — gradient boosting with exogenous features - Facebook Prophet — decomposable trend-seasonality model All models were evaluated against a naive baseline across one-month, two-month and three-month forecast horizons. --- ## 📊 Results Summary ### Overall Model Comparison — Tier 1 (Adamawa, Borno, Yobe) | Model | Combinations Won | Win Rate | Average MAPE | |:------|:---:|:---:|:---:| | Prophet | 10 of 13 | 76.9% | 30.8% | | XGBoost | 2 of 13 | 15.4% | 49.5% | | SARIMA | 1 of 13 | 7.7% | 53.6% | ### Prophet Accuracy by Forecast Horizon | Horizon | Prophet Avg MAPE | XGBoost Avg MAPE | |:--------|:---:|:---:| | H1 — 1 month ahead | 24.5% | 51.1% | | H2 — 2 months ahead | 28.5% | …