dez-data/ghana-gdp-forecasting

Time series forecasting of Ghana's GDP using ARIMA, ETS, and Prophet models in R # A Comparative Analysis: ARIMA vs ETS vs Prophet in Forecasting Ghana's GDP (1960 - 2023) Time series forecasting analysis comparing ARIMA, ETS, and Prophet models using 63 years of Ghana's GDP data (1960-2023). **View Full Analysis Report (HTML)** --- ## Project Overview This project compares three time series forecasting approaches: - **ARIMA(1,2,3)** - Auto-Regressive Integrated Moving Average - **ETS** - Exponential Smoothing State Space Model - **Prophet** - Facebook's forecasting algorithm **Key Finding:** Prophet outperformed ARIMA by 27% and ETS by 37% based on RMSE evaluation, making it the optimal model for Ghana's GDP forecasting. --- ## Results Summary ### Model Performance | Model | RMSE | MAE | Performance vs Prophet | |-------|------|-----|------------------------| | Prophet | 6,699 | 6,719 | Best (Baseline) | | ARIMA(1,2,3) | 9,146 | 9,185 | -27% worse | | ETS | 10,644 | 10,702 | -37% worse | ### Key Visualizations **Ghana's GDP Historical Trend (1960-2023)** .png) **ARIMA Forecast** **ETS Forecast** **Prophet Forecast** **Model Performance Comparison - RMSE** **Model Performance Comparison - MAE** --- ## Key Insights 1. **Prophet's Superior Performance:** - 27% lower error rate than ARIMA (RMSE: 6,699 vs 9,146) - 37% lower error rate than ETS (RMSE: 6,699 vs 10,644) - Consistent advantage across both RMSE and MAE metrics 2. **Why Prophet Won:** - Better handling of structural breaks in Ghana's economic history (e.g., 1983 economic crisis, 2008 financial crisis) - Automatic changepoint detection adapts to regime changes - Robust to outliers and missing data 3. **ARIMA Limitation:** - Assumes stationary patterns after differencing (d=2 required) - Struggles with Ghana's volatile economic trajectory and multiple structural changes - Best suited for more stable economic environments 4. **ETS Performance:** - Simple exponential smoothing insuffic …