Tools › Statistics › Forecasting › Forecasting › Example
FPP2 §7.1: simple exponential smoothing of Saudi Arabian oil production 1996–2013
| Tool | Forecasting |
| Source | Hyndman, R. J., and Athanasopoulos, G. (2018). Forecasting: Principles and Practice, 2nd ed. OTexts, Melbourne. Section 7.1, Example "Oil production", Table 7.2 (copy in References/Forecasting-FPP/). |
| Level | Undergraduate |
Problem
Annual oil production in Saudi Arabia (millions of tonnes) from 1996 to 2013 shows no clear trend or seasonality. Fit simple exponential smoothing with the smoothing parameter and the initial level estimated by least squares, forecast the next five years and report the training accuracy.
Inputs to Enter
- Series (18 values, 1996–2013): 445.36 453.20 454.41 422.38 456.04 440.39 425.19 486.21 500.43 521.28 508.95 488.89 509.87 456.72 473.82 525.95 549.83 542.34
- Method: Simple exponential smoothing; season length 1; horizon 5; level α left blank (fitted); confidence 95 %
Answers
- α = 0.8339, ℓ₀ = 446.6
- Forecast 542.68 for every horizon (flat)
- Training accuracy: ME 6.4, RMSE 28.12, MAE 22.26, MAPE 4.61, MASE 0.93
- Level at 1999 (ℓ₄) from the printed table: 427.63
Notes
Paste the 18 values, choose SES with season length 1 and leave α blank. Forecast reproduced within 0.02 %, α within 0.005 (test_forecasting.php). Prediction intervals use the ETS(A,N,N) multiplier 1 + (h − 1)α² of FPP2 §7.5.