lyna1908/Algeria-Oil-Production-TSAC

ARIMA time series forecasting of Algeria's crude oil production — R, Box-Jenkins methodology, MAPE 7.4% # Time Series Analysis of Crude Oil Production in Algeria (2000–2025) ## Overview A complete Box-Jenkins ARIMA analysis of Algeria's monthly crude oil production from January 2000 to December 2024 (300 observations). ## Key Results - **Model:** ARIMA(0,1,1) with Box-Cox transformation (λ=2) - **Forecast accuracy:** MAPE = 7.4% on withheld 2024 data - **Finding:** Production declined ~34% from its 2008 peak (1,710 Mb/d) to 1,130 Mb/d by end-2024 - **2025 Forecast:** ~638 Mb/d average production ## Methodology (Box-Jenkins Pipeline) 1. Exploratory Data Analysis (time plot, ACF, descriptive stats) 2. Deterministic trend analysis (linear + quadratic) 3. Variance stabilisation (Box-Cox, λ=2) 4. Stationarity testing (ADF test → d=1 differencing) 5. Model specification (ACF/PACF + AIC/BIC grid search) 6. Parameter estimation (Maximum Likelihood) 7. Residual diagnostics (Ljung-Box, Shapiro-Wilk, Runs test, ACF) 8. Forecasting (12-month ahead, with 95% prediction intervals) ## Data Source U.S. Energy Information Administration (EIA) Monthly crude oil production — Algeria (Thousand Barrels/Day) Period: January 2000 – December 2024 eia.gov ## Tools & Libraries - **Language:** R - **Packages:** `forecast`, `tseries`, `MASS`, `ggplot2`, `lmtest`, `knitr` - **Environment:** Google Colab / Jupyter with R kernel ## Repository Structure ``` ├── algeria_oil_clean.csv # Cleaned monthly production data ├── TSAC_Algeria_Oil.ipynb # Full analysis notebook └── README.md ``` ## Context This project was completed as part of the **TSAC (Time Series Analysis and Classification)** course at **ENSIA**, 2025/2026.