An end-to-end time-series forecasting project addressing an environmental-health problem
# Nigeria AQI Forecasting
## 12-Month Air Quality Index Forecast Using SARIMA
An end-to-end data science project that analyzes historical air-quality data from Nigerian cities and forecasts monthly Air Quality Index (AQI) for 2026.
---
## Project Overview
Air-quality monitoring often focuses on describing current or historical pollution levels. This project moves from **descriptive analytics to predictive analytics** by using historical monthly AQI patterns to forecast future air-quality conditions.
The project uses data covering **2014–2025**, performs exploratory data analysis and time-series diagnostics, builds a seasonal SARIMA model, evaluates it on an unseen 2025 test period, and produces a 12-month AQI forecast for 2026.
### Research Question
**Can historical monthly AQI patterns be used to forecast Nigeria's air quality for the next 12 months?**
---
## Objectives
- Understand the distribution and trends of AQI in the dataset.
- Identify monthly/annual seasonal patterns.
- Compare AQI levels across Nigerian cities.
- Investigate relationships between AQI and major pollutants.
- Build and validate a seasonal time-series forecasting model.
- Forecast monthly AQI for January–December 2026.
- Identify limitations and opportunities for future improvement.
---
## Dataset
The analysis dataset contains:
- **10,368 observations**
- **72 cities**
- **37 states**
- **2014–2025**
- Monthly air-quality observations
- AQI plus multiple pollutant measurements
Key pollutant variables include PM2.5, PM10, NO, NO2, NOx, NH3, CO, SO2, O3, Benzene, Toluene and Xylene.
### Data availability
The raw CSV is **included in this repository**.
```text
nigeria_air_quality_2014_2025.csv
```
---
## Methodology
```text
Raw Data
↓
Data Cleaning & Validation
↓
Exploratory Data Analysis
↓
Monthly Time-Series Construction
↓
Seasonal Decomposition
↓
Stationarity Testing
↓
ACF/PACF Analysis
↓
Chronological Train/Test Split
↓
SARIMA Modelling
↓
2025 Holdout Validation
↓
Res …