Trend/Seasonal Models w. 1st Order Exp.

Smoothing
The following formula is used in forecast strategies 20, 21, 30, 31, 40 and 41, and in forecast strategies 50 to 56
where a trend, seasonal, or seasonal trend model is determined. The calculation takes into account both trend
and seasonal variations. The basic value, the trend value and the seasonal index are calculated after the initial
period.
Formula for First-Order Exponential Smoothing in a Trend, Seasonal or Seasonal Trend Model

Forecast value for the perio d ( t +1 ) :

P (t +1 )= [ G ( t )+ i∙ T ( t ) ] ∙ S (t−L+i )

Basic value:G ( t )=G ( t−1 )+ T ( t−1 ) +α ∙

[ V (t)
S (t−L)
−G ( t−1 )−T (t−1)
]
Trend value :T ( t )=T ( t −1 )+ β ∙ [ G ( t )−G ( t−1 ) +T (t−1) ]

Seasonal index : S ( t ) =S ( t−L )+ γ ∙ [ V (t)

G (t)
−S (t−L) ]
P (t +i )=the forecast calculated for the period ( t +i ) ∈the current period ( t ) i=forecast horizon
G (t )=thecurrent basic value for thecurrent period ( t )
G (t−1 )=the previous basic value ¿ the previous period L= period lenght ( often 12 )
V ( t )=actual demand ( history ) for the current period ( t )
T ( t )=the current trend value calculated for the current period
T ( t−1 )=the previous trend value ¿ the previous period S ( t ) =the seasonal index for the period ( t )
'
S ( t−L )=the previous seasonal index for the period ( t ) α =smoothing factor for the basic value ' G , 0< α <1
'
β=smoothing factor for the trend value' T ,0< β<1
'
γ =smoothing factor for the seasonal indices ' S , 0< γ <1

 Constant model: T ( t )=0, β=0, S ( t ) =1,0, γ=0

 Trend model: S ( t ) =1,0 γ =0
 Seasonal model: T ( t )=0, β=0
Use the trend, seasonal or seasonal trend model with first-order exponential smoothing for time series that
have trend-like patterns and/or seasonal variations.