지수 평활법(Exponential smoothing)

single exponential smoothing 

 

future expected value = past real value * a + past predicted value * (1-a)

 

• Forecasting Formula:
  • Ft+1F_{t+1}Ft+1​: Forecast for the next period.
  • XtX_tXt​: Actual value at time ttt.
  • FtF_tFt​: Forecast value at time ttt.
  • α\alphaα: Smoothing constant (0 ≤ α\alphaα ≤ 1).
• Smoothing Constant (α\alphaα):
  • Determines the weight of recent observations versus older ones.
  • A higher α\alphaα (closer to 1) gives more weight to recent data, making the model more responsive to changes.
  • A lower α\alphaα (closer to 0) gives more weight to older data, resulting in a smoother forecast.
• Initialization:
  • The initial forecast (F1F_1F1​) can be set to the first data point (X1X_1X1​) or the average of the first few observations.
• Exponential Decay:
  • The method assigns exponentially decreasing weights to past observations, ensuring that recent observations influence the forecast more than older ones.

 

Excel usage
= FORECAST.ETS(target_data, values, timeline, seasonal period, null data management, repeated timeline management)