Rob J Hyndman

Forecasting using

5. Exponential smoothing methods

[Link]/fpp/7/

Forecasting using R 1
Outline

1 Simple exponential smoothing

2 Non-seasonal trend methods

Forecasting using R Simple exponential smoothing 2

Simple methods
Random walk forecasts
yT +1|T = yT

Average forecasts
T
1X
yT +1|T = yt
T
t =1

Want something in between that weights most

recent data more highly.
Simple exponential smoothing uses a weighted
moving average with weights that decrease
exponentially.
Forecasting using R Simple exponential smoothing 3
Simple methods
Random walk forecasts
yT +1|T = yT

Average forecasts
T
1X
yT +1|T = yt
T
t =1

Want something in between that weights most

recent data more highly.
Simple exponential smoothing uses a weighted
moving average with weights that decrease
exponentially.
Forecasting using R Simple exponential smoothing 3
Simple methods
Random walk forecasts
yT +1|T = yT

Average forecasts
T
1X
yT +1|T = yt
T
t =1

Want something in between that weights most

recent data more highly.
Simple exponential smoothing uses a weighted
moving average with weights that decrease
exponentially.
Forecasting using R Simple exponential smoothing 3
Simple methods
Random walk forecasts
yT +1|T = yT

Average forecasts
T
1X
yT +1|T = yt
T
t =1

Want something in between that weights most

recent data more highly.
Simple exponential smoothing uses a weighted
moving average with weights that decrease
exponentially.
Forecasting using R Simple exponential smoothing 3
Simple Exponential Smoothing
Forecast equation
yT +1|T = yT + (1 )yT 1 + (1 )2 yT 2 + ,

where 0 1.

Weights assigned to observations for:

Observation = 0.2 = 0.4 = 0.6 = 0.8

yT 0.2 0.4 0.6 0.8

yT 1 0.16 0.24 0.24 0.16
yT 2 0.128 0.144 0.096 0.032
yT 3 0.1024 0.0864 0.0384 0.0064
yT 4 (0.2)(0.8)4 (0.4)(0.6)4 (0.6)(0.4)4 (0.8)(0.2)4
yT 5 (0.2)(0.8)5 (0.4)(0.6)5 (0.6)(0.4)5 (0.8)(0.2)5

Forecasting using R Simple exponential smoothing 4

Simple Exponential Smoothing
Forecast equation
yT +1|T = yT + (1 )yT 1 + (1 )2 yT 2 + ,

where 0 1.

Weights assigned to observations for:

Observation = 0.2 = 0.4 = 0.6 = 0.8

yT 0.2 0.4 0.6 0.8

yT 1 0.16 0.24 0.24 0.16
yT 2 0.128 0.144 0.096 0.032
yT 3 0.1024 0.0864 0.0384 0.0064
yT 4 (0.2)(0.8)4 (0.4)(0.6)4 (0.6)(0.4)4 (0.8)(0.2)4
yT 5 (0.2)(0.8)5 (0.4)(0.6)5 (0.6)(0.4)5 (0.8)(0.2)5

Forecasting using R Simple exponential smoothing 4

Simple Exponential Smoothing
Weighted average form

yt+1|t = yt + (1 )yt|t1

for t = 1, . . . , T, where 0 1 is the smoothing

parameter.
The process has to start somewhere, so we let the
first forecast of y1 be denoted by `0 . Then
y2|1 = y1 + (1 )`0
y3|2 = y2 + (1 )y2|1
y4|3 = y3 + (1 )y3|2
..
.
Forecasting using R Simple exponential smoothing 5
Simple Exponential Smoothing
Weighted average form

yt+1|t = yt + (1 )yt|t1

for t = 1, . . . , T, where 0 1 is the smoothing

parameter.
The process has to start somewhere, so we let the
first forecast of y1 be denoted by `0 . Then
y2|1 = y1 + (1 )`0
y3|2 = y2 + (1 )y2|1
y4|3 = y3 + (1 )y3|2
..
.
Forecasting using R Simple exponential smoothing 5
Simple Exponential Smoothing
yt+1|t = yt + (1 )yt|t1
Substituting each equation into the following equation:
y3|2 = y2 + (1 )y2|1
= y2 + (1 ) [y1 + (1 )`0 ]
= y2 + (1 )y1 + (1 )2 `0
y4|3 = y3 + (1 )[y2 + (1 )y1 + (1 )2 `0 ]
= y3 + (1 )y2 + (1 )2 y1 + (1 )3 `0
..
.
yT +1|T = yT + (1 )yT 1 + (1 )2 yT 2 + + (1 )T `0

Exponentially weighted average

T 1
X
yT +1|T = (1 )j yT j + (1 )T `0
j=0
Forecasting using R Simple exponential smoothing 6
Simple Exponential Smoothing
yt+1|t = yt + (1 )yt|t1
Substituting each equation into the following equation:
y3|2 = y2 + (1 )y2|1
= y2 + (1 ) [y1 + (1 )`0 ]
= y2 + (1 )y1 + (1 )2 `0
y4|3 = y3 + (1 )[y2 + (1 )y1 + (1 )2 `0 ]
= y3 + (1 )y2 + (1 )2 y1 + (1 )3 `0
..
.
yT +1|T = yT + (1 )yT 1 + (1 )2 yT 2 + + (1 )T `0

Exponentially weighted average

T 1
X
yT +1|T = (1 )j yT j + (1 )T `0
j=0
Forecasting using R Simple exponential smoothing 6
Simple exponential smoothing

Initialization
Last term in weighted moving average is
(1 )T `0 .
So value of `0 plays a role in all subsequent
forecasts.
Weight is small unless close to zero or T
small.
Common to set `0 = y1 . Better to treat it as a
parameter, along with .

Forecasting using R Simple exponential smoothing 7

Simple exponential smoothing

Initialization
Last term in weighted moving average is
(1 )T `0 .
So value of `0 plays a role in all subsequent
forecasts.
Weight is small unless close to zero or T
small.
Common to set `0 = y1 . Better to treat it as a
parameter, along with .

Forecasting using R Simple exponential smoothing 7

Simple exponential smoothing

Initialization
Last term in weighted moving average is
(1 )T `0 .
So value of `0 plays a role in all subsequent
forecasts.
Weight is small unless close to zero or T
small.
Common to set `0 = y1 . Better to treat it as a
parameter, along with .

Forecasting using R Simple exponential smoothing 7

Simple exponential smoothing

Initialization
Last term in weighted moving average is
(1 )T `0 .
So value of `0 plays a role in all subsequent
forecasts.
Weight is small unless close to zero or T
small.
Common to set `0 = y1 . Better to treat it as a
parameter, along with .

Forecasting using R Simple exponential smoothing 7

 Simple exponential smoothing
6000
5500
No. strikes in US

5000
4500
4000
3500

1950 1960 1970 1980 1990

Year

Forecasting using R Simple exponential smoothing 8

 Simple exponential smoothing
6000

= 0.01
5500
No. strikes in US

5000
4500
4000
3500

1950 1960 1970 1980 1990

Year

Forecasting using R Simple exponential smoothing 9

Simple exponential smoothing

Optimization
We can choose and `0 by minimizing MSE:
T
1 X
MSE = (yt yt|t1 )2
T1
t =2

Unlike regression there is no closed form

solution use numerical optimization.

Forecasting using R Simple exponential smoothing 10

Simple exponential smoothing

Optimization
We can choose and `0 by minimizing MSE:
T
1 X
MSE = (yt yt|t1 )2
T1
t =2

Unlike regression there is no closed form

solution use numerical optimization.

Forecasting using R Simple exponential smoothing 10

 Simple exponential smoothing
6000

= 0.01

MSE= 1976007
5500
No. strikes in US

5000
4500
4000
3500

1950 1960 1970 1980 1990

Year

Forecasting using R Simple exponential smoothing 11

 Simple exponential smoothing
2.0
1.8
1.6
MSE ('000 000)

1.4
1.2

= 0.68
1.0
0.8

0.0 0.2 0.4 0.6 0.8 1.0

alpha

Forecasting using R Simple exponential smoothing 12

Simple exponential smoothing

Multi-step forecasts

yT +h|T = yT +1|T , h = 2, 3, . . .

A flat forecast function.

Remember, a forecast is an estimated mean of
a future value.
So with no trend, no seasonality, and no other
patterns, the forecasts are constant.

Forecasting using R Simple exponential smoothing 13

Simple exponential smoothing

Multi-step forecasts

yT +h|T = yT +1|T , h = 2, 3, . . .

A flat forecast function.

Remember, a forecast is an estimated mean of
a future value.
So with no trend, no seasonality, and no other
patterns, the forecasts are constant.

Forecasting using R Simple exponential smoothing 13

Simple exponential smoothing

Multi-step forecasts

yT +h|T = yT +1|T , h = 2, 3, . . .

A flat forecast function.

Remember, a forecast is an estimated mean of
a future value.
So with no trend, no seasonality, and no other
patterns, the forecasts are constant.

Forecasting using R Simple exponential smoothing 13

Simple exponential smoothing

Multi-step forecasts

yT +h|T = yT +1|T , h = 2, 3, . . .

A flat forecast function.

Remember, a forecast is an estimated mean of
a future value.
So with no trend, no seasonality, and no other
patterns, the forecasts are constant.

Forecasting using R Simple exponential smoothing 13

Example: Oil production
Time Observed = 0.2 = 0.6 = 0.89
Year period t values yt Level `t

0 446.7 446.7 447.5

1996 1 446.7 446.7 446.7 446.7
1997 2 454.5 448.2 450.6 453.6
1998 3 455.7 449.7 453.1 455.4
1999 4 423.6 444.5 438.4 427.1
2000 5 456.3 446.8 447.3 453.1
2001 6 440.6 445.6 444.0 441.9
2002 7 425.3 441.5 434.6 427.1
2003 8 485.1 450.3 459.9 478.9
2004 9 506.0 461.4 483.0 503.1
2005 10 526.8 474.5 504.9 524.2
2006 11 514.3 482.5 509.6 515.3
2007 12 494.2 484.8 501.9 496.5
h Forecasts yT +h|T
2008 1 484.8 501.9 496.5
2009 2 484.8 501.9 496.5
2010 3 484.8 501.9 496.5
= 0.89 and `0 = 447.5 are obtained by minimising SSE over periods t = 1, 2, . . . , 12.
Forecasting using R Simple exponential smoothing 14
 Example: Oil production

data
520

= 0.2

= 0.6

= 0.89

500

Oil (millions of tonnes)

480

460

440

420

1996 1998 2000 2002 2004 2006 2008 2010

Year

Forecasting using R Simple exponential smoothing 15

SES in R

fit1 <- ses(oildata, alpha=0.2,

initial="simple", h=3)
fit2 <- ses(oildata, alpha=0.6,
initial="simple", h=3)
fit3 <- ses(oildata, h=3)

accuracy(fit1)
accuracy(fit2)
accuracy(fit3)

Forecasting using R Simple exponential smoothing 16

Equivalent forms
Weighted average form
yt+1|t = yt + (1 )yt|t1

Error correction form

yt+1|t = yt|t1 + (yt yt|t1 )

Component form
yt+h|t = `t
`t = yt + (1 )`t1

`t = estimate of level of series.

Forecasting using R Simple exponential smoothing 17
Equivalent forms
Weighted average form
yt+1|t = yt + (1 )yt|t1

Error correction form

yt+1|t = yt|t1 + (yt yt|t1 )

Component form
yt+h|t = `t
`t = yt + (1 )`t1

`t = estimate of level of series.

Forecasting using R Simple exponential smoothing 17
Equivalent forms
Weighted average form
yt+1|t = yt + (1 )yt|t1

Error correction form

yt+1|t = yt|t1 + (yt yt|t1 )

Component form
yt+h|t = `t
`t = yt + (1 )`t1

`t = estimate of level of series.

Forecasting using R Simple exponential smoothing 17
Outline

1 Simple exponential smoothing

2 Non-seasonal trend methods

Forecasting using R Non-seasonal trend methods 18

Holts local trend method
Holt (1957) extended SES to allow forecasting
of data with trends.
Two smoothing parameters: and (with
values between 0 and 1).
yt+h|t = `t + hbt
`t = yt + (1 )(`t1 + bt1 )
bt = (`t `t1 ) + (1 )bt1

`t denotes an estimate of the level of the series

at time t
bt denotes an estimate of the slope of the
series at time t.
Forecasting using R Non-seasonal trend methods 19
Holts local trend method
Holt (1957) extended SES to allow forecasting
of data with trends.
Two smoothing parameters: and (with
values between 0 and 1).
yt+h|t = `t + hbt
`t = yt + (1 )(`t1 + bt1 )
bt = (`t `t1 ) + (1 )bt1

`t denotes an estimate of the level of the series

at time t
bt denotes an estimate of the slope of the
series at time t.
Forecasting using R Non-seasonal trend methods 19
Holts local trend method
Holt (1957) extended SES to allow forecasting
of data with trends.
Two smoothing parameters: and (with
values between 0 and 1).
yt+h|t = `t + hbt
`t = yt + (1 )(`t1 + bt1 )
bt = (`t `t1 ) + (1 )bt1

`t denotes an estimate of the level of the series

at time t
bt denotes an estimate of the slope of the
series at time t.
Forecasting using R Non-seasonal trend methods 19
Holts local trend method
Holt (1957) extended SES to allow forecasting
of data with trends.
Two smoothing parameters: and (with
values between 0 and 1).
yt+h|t = `t + hbt
`t = yt + (1 )(`t1 + bt1 )
bt = (`t `t1 ) + (1 )bt1

`t denotes an estimate of the level of the series

at time t
bt denotes an estimate of the slope of the
series at time t.
Forecasting using R Non-seasonal trend methods 19
Holts local trend method
Holt (1957) extended SES to allow forecasting
of data with trends.
Two smoothing parameters: and (with
values between 0 and 1).
yt+h|t = `t + hbt
`t = yt + (1 )(`t1 + bt1 )
bt = (`t `t1 ) + (1 )bt1

`t denotes an estimate of the level of the series

at time t
bt denotes an estimate of the slope of the
series at time t.
Forecasting using R Non-seasonal trend methods 19
Holts local trend method
Holt (1957) extended SES to allow forecasting
of data with trends.
Two smoothing parameters: and (with
values between 0 and 1).
yt+h|t = `t + hbt
`t = yt + (1 )(`t1 + bt1 )
bt = (`t `t1 ) + (1 )bt1

`t denotes an estimate of the level of the series

at time t
bt denotes an estimate of the slope of the
series at time t.
Forecasting using R Non-seasonal trend methods 19
Holts local trend method

Optimization
Need to find and by minimizing the value
of MSE.
We also optimize MSE for `0 and b0 .
Optimizing in four dimensions is getting tricky!

Forecasting using R Non-seasonal trend methods 20

Holts method in R
fit1 <- holt(strikes)
plot(fit1$model)
plot(fit1, [Link]=FALSE)
lines(fitted(fit1), col="red")
fit1$model

fit2 <- ses(strikes)

plot(fit2$model)
plot(fit2, [Link]=FALSE)
lines(fit1$mean, col="red")

accuracy(fit1)
accuracy(fit2)

Forecasting using R Non-seasonal trend methods 21

Comparing Holt and SES

Holts method will almost always have better

in-sample RMSE because it is optimized over
one additional parameter.
It may not be better on other measures.
You need to compare out-of-sample RMSE
(using a test set) for the comparison to be
useful.
But we dont have enough data.
A better method for comparison will be in the
next session!

Forecasting using R Non-seasonal trend methods 22

Comparing Holt and SES

Holts method will almost always have better

in-sample RMSE because it is optimized over
one additional parameter.
It may not be better on other measures.
You need to compare out-of-sample RMSE
(using a test set) for the comparison to be
useful.
But we dont have enough data.
A better method for comparison will be in the
next session!

Forecasting using R Non-seasonal trend methods 22

Comparing Holt and SES

Holts method will almost always have better

in-sample RMSE because it is optimized over
one additional parameter.
It may not be better on other measures.
You need to compare out-of-sample RMSE
(using a test set) for the comparison to be
useful.
But we dont have enough data.
A better method for comparison will be in the
next session!

Forecasting using R Non-seasonal trend methods 22

Comparing Holt and SES

Holts method will almost always have better

in-sample RMSE because it is optimized over
one additional parameter.
It may not be better on other measures.
You need to compare out-of-sample RMSE
(using a test set) for the comparison to be
useful.
But we dont have enough data.
A better method for comparison will be in the
next session!

Forecasting using R Non-seasonal trend methods 22

Comparing Holt and SES

Holts method will almost always have better

in-sample RMSE because it is optimized over
one additional parameter.
It may not be better on other measures.
You need to compare out-of-sample RMSE
(using a test set) for the comparison to be
useful.
But we dont have enough data.
A better method for comparison will be in the
next session!

Forecasting using R Non-seasonal trend methods 22

Exponential trend method
Multiplicative version of Holts method
yt+h|t = `t bht
`t = yt + (1 )(`t1 bt1 )
bt = (`t /`t1 ) + (1 )bt1

`t denotes an estimate of the level of the series at

time t
bt denotes an estimate of the relative growth of the
series at time t.
In R: holt(x, exponential=TRUE)
Comparing additive and multiplicative trend
methods in-sample is ok because they have the
same number of parameters to optimize.
Forecasting using R Non-seasonal trend methods 23
Exponential trend method
Multiplicative version of Holts method
yt+h|t = `t bht
`t = yt + (1 )(`t1 bt1 )
bt = (`t /`t1 ) + (1 )bt1

`t denotes an estimate of the level of the series at

time t
bt denotes an estimate of the relative growth of the
series at time t.
In R: holt(x, exponential=TRUE)
Comparing additive and multiplicative trend
methods in-sample is ok because they have the
same number of parameters to optimize.
Forecasting using R Non-seasonal trend methods 23
Exponential trend method
Multiplicative version of Holts method
yt+h|t = `t bht
`t = yt + (1 )(`t1 bt1 )
bt = (`t /`t1 ) + (1 )bt1

`t denotes an estimate of the level of the series at

time t
bt denotes an estimate of the relative growth of the
series at time t.
In R: holt(x, exponential=TRUE)
Comparing additive and multiplicative trend
methods in-sample is ok because they have the
same number of parameters to optimize.
Forecasting using R Non-seasonal trend methods 23
Exponential trend method
Multiplicative version of Holts method
yt+h|t = `t bht
`t = yt + (1 )(`t1 bt1 )
bt = (`t /`t1 ) + (1 )bt1

`t denotes an estimate of the level of the series at

time t
bt denotes an estimate of the relative growth of the
series at time t.
In R: holt(x, exponential=TRUE)
Comparing additive and multiplicative trend
methods in-sample is ok because they have the
same number of parameters to optimize.
Forecasting using R Non-seasonal trend methods 23
Exponential trend method
Multiplicative version of Holts method
yt+h|t = `t bht
`t = yt + (1 )(`t1 bt1 )
bt = (`t /`t1 ) + (1 )bt1

`t denotes an estimate of the level of the series at

time t
bt denotes an estimate of the relative growth of the
series at time t.
In R: holt(x, exponential=TRUE)
Comparing additive and multiplicative trend
methods in-sample is ok because they have the
same number of parameters to optimize.
Forecasting using R Non-seasonal trend methods 23
Damped trend method
Gardner and McKenzie (1985) suggested that
the trends should be damped to be more
conservative for longer forecast horizons.
Two smoothing parameters: and (with
values between 0 and 1), and one damping
parameter 0 < < 1.
yt+h|t = `t + ( + 2 + + h1 )bt
`t = yt + (1 )(`t1 + bt1 )
bt = (`t `t1 ) + (1 )bt1

dampens the trend so it approaches a

constant.
Forecasting using R Non-seasonal trend methods 24
Damped trend method
Gardner and McKenzie (1985) suggested that
the trends should be damped to be more
conservative for longer forecast horizons.
Two smoothing parameters: and (with
values between 0 and 1), and one damping
parameter 0 < < 1.
yt+h|t = `t + ( + 2 + + h1 )bt
`t = yt + (1 )(`t1 + bt1 )
bt = (`t `t1 ) + (1 )bt1

dampens the trend so it approaches a

constant.
Forecasting using R Non-seasonal trend methods 24
Damped trend method
Gardner and McKenzie (1985) suggested that
the trends should be damped to be more
conservative for longer forecast horizons.
Two smoothing parameters: and (with
values between 0 and 1), and one damping
parameter 0 < < 1.
yt+h|t = `t + ( + 2 + + h1 )bt
`t = yt + (1 )(`t1 + bt1 )
bt = (`t `t1 ) + (1 )bt1

dampens the trend so it approaches a

constant.
Forecasting using R Non-seasonal trend methods 24
Damped trend method
Gardner and McKenzie (1985) suggested that
the trends should be damped to be more
conservative for longer forecast horizons.
Two smoothing parameters: and (with
values between 0 and 1), and one damping
parameter 0 < < 1.
yt+h|t = `t + ( + 2 + + h1 )bt
`t = yt + (1 )(`t1 + bt1 )
bt = (`t `t1 ) + (1 )bt1

dampens the trend so it approaches a

constant.
Forecasting using R Non-seasonal trend methods 24
Damped trend method
Gardner and McKenzie (1985) suggested that
the trends should be damped to be more
conservative for longer forecast horizons.
Two smoothing parameters: and (with
values between 0 and 1), and one damping
parameter 0 < < 1.
yt+h|t = `t + ( + 2 + + h1 )bt
`t = yt + (1 )(`t1 + bt1 )
bt = (`t `t1 ) + (1 )bt1

dampens the trend so it approaches a

constant.
Forecasting using R Non-seasonal trend methods 24
Damped trend method
Forecasts from damped Holt's method
5500
4500
3500
2500

1950 1960 1970 1980 1990

Forecasting using R Non-seasonal trend methods 25

Damped trend method
yt+h|t = `t + ( + 2 + + h1 )bt
`t = yt + (1 )(`t1 + bt1 )
bt = (`t `t1 ) + (1 )bt1

If = 1, this is the same as Holts method.

can be estimated along with and by
minimizing the MSE.
Damped trend method often gives better
forecasts than linear trend.
Forecasts converge to `T + bT /(1 ) as
h .
Forecasting using R Non-seasonal trend methods 26
Damped trend method
yt+h|t = `t + ( + 2 + + h1 )bt
`t = yt + (1 )(`t1 + bt1 )
bt = (`t `t1 ) + (1 )bt1

If = 1, this is the same as Holts method.

can be estimated along with and by
minimizing the MSE.
Damped trend method often gives better
forecasts than linear trend.
Forecasts converge to `T + bT /(1 ) as
h .
Forecasting using R Non-seasonal trend methods 26
Damped trend method
yt+h|t = `t + ( + 2 + + h1 )bt
`t = yt + (1 )(`t1 + bt1 )
bt = (`t `t1 ) + (1 )bt1

If = 1, this is the same as Holts method.

can be estimated along with and by
minimizing the MSE.
Damped trend method often gives better
forecasts than linear trend.
Forecasts converge to `T + bT /(1 ) as
h .
Forecasting using R Non-seasonal trend methods 26
Damped trend method
yt+h|t = `t + ( + 2 + + h1 )bt
`t = yt + (1 )(`t1 + bt1 )
bt = (`t `t1 ) + (1 )bt1

If = 1, this is the same as Holts method.

can be estimated along with and by
minimizing the MSE.
Damped trend method often gives better
forecasts than linear trend.
Forecasts converge to `T + bT /(1 ) as
h .
Forecasting using R Non-seasonal trend methods 26
Damped trend method
yt+h|t = `t + ( + 2 + + h1 )bt
`t = yt + (1 )(`t1 + bt1 )
bt = (`t `t1 ) + (1 )bt1

If = 1, this is the same as Holts method.

can be estimated along with and by
minimizing the MSE.
Damped trend method often gives better
forecasts than linear trend.
Forecasts converge to `T + bT /(1 ) as
h .
Forecasting using R Non-seasonal trend methods 26
Multiplicative damped trend method

Taylor (2003) introduced multiplicative damping.

(+2 ++h )
yt+h|t = `t bt
`t = yt + (1 )(`t1 bt1 )

bt = (`t /`t1 ) + (1 )bt1

= 1 gives exponential trend method

/(1)
Forecasts converge to `T + bT as h .

Forecasting using R Non-seasonal trend methods 27

Multiplicative damped trend method

Taylor (2003) introduced multiplicative damping.

(+2 ++h )
yt+h|t = `t bt
`t = yt + (1 )(`t1 bt1 )

bt = (`t /`t1 ) + (1 )bt1

= 1 gives exponential trend method

/(1)
Forecasts converge to `T + bT as h .

Forecasting using R Non-seasonal trend methods 27

Multiplicative damped trend method

Taylor (2003) introduced multiplicative damping.

(+2 ++h )
yt+h|t = `t bt
`t = yt + (1 )(`t1 bt1 )

bt = (`t /`t1 ) + (1 )bt1

= 1 gives exponential trend method

/(1)
Forecasts converge to `T + bT as h .

Forecasting using R Non-seasonal trend methods 27