TitilayoAkinloye/nigeria-electricity-sarima-forecast

SARIMA time series forecasting of electricity distribution in Nigeria — BSc Statistics dissertation # Time Series Analysis of Electricity Distribution in Nigeria ## Overview This project applies time series forecasting to electricity distribution data in Nigeria, using the Box-Jenkins methodology to build a Seasonal ARIMA model capable of predicting future distribution patterns. The model was developed as a final year dissertation in Statistics at the Federal University of Agriculture, Abeokuta (FUNAAB). Electricity distribution in Nigeria is notoriously volatile — supply shortages, seasonal demand shifts, and infrastructure constraints all make it difficult to plan around. This project explores whether a statistical model can bring some predictability to that volatility. ## Methodology The analysis followed the **Box-Jenkins methodology**, a structured approach to time series modeling with three main stages: 1. **Identification** — analyzing 120 monthly observations to determine the appropriate model structure, including tests for stationarity and seasonality 2. **Estimation** — fitting a **SARIMA(2,1,1)(2,0,0)[12]** model (Seasonal Autoregressive Integrated Moving Average) to the data 3. **Diagnostic checking** — validating the model through residual analysis to confirm it captures the underlying patterns without leaving unexplained structure behind The model was then used to generate a **60-month forecast**, projecting electricity distribution trends five years into the future. **Tools used:** R (time series analysis, model fitting, forecasting) ## Key Findings - Nigeria's electricity distribution showed a statistically significant upward trend from 2015–2024, alongside a strong, stable seasonal pattern — supply consistently peaks in December/January and dips in February/May. - The best-fit model, SARIMA(2,1,1)(2,0,0)[12], outperformed a naive seasonal forecast by 52% (MASE = 0.48) with a MAPE of 4.87% — placing it in the "good accuracy" range for time series forecasting. - The 5-year forecast (2025–2029) projects distribution stabilizing at ~1,956–1,968 …