Exponential Smoothing Equations

Exponential smoothing was initially introduced in the statistical literature without considering the past work done by Robert Goodell Brown in 1956 and then further expanded by Charles C. Holt in 1957. Exponential smoothing is a reliable principle for smoothing time series data through the exponential window function. The controlling input of the exponential smoothing calculation is stated as the smoothing factor or the smoothing constant.

Forecast of the weighted averages of past observations are introduced using exponential smoothing methods, with the weights breaking down exponentially as the observations get formed. In other words, the more the latest the observation the higher the corresponding weight.

As we are aware of the simple moving average the past observations are weighted equally, exponential functions are used to assign exponentially decreasing weights over time. It can be easily applied for making determinations on the basis of prior assumptions by the user, such as seasonality. Exponential smoothing is primarily used for time-series data analysis.

Exponential Smoothing Formula

The exponential smoothing formula is derived by:

st = θxt+(1 – θ)st-1= st-1+ θ(xt – st-1)

Here,

st is a former smoothed statistic, it is the simple weighted average of present observation xt

st-1 is former smoothed statistic

θ is smoothing factor of data; 0 < θ < 1

t is time period

If the value of the smoothing factor is greater, then the level of smoothing will be minimized. Value of α nearer to 1 minimum smoothing effect and offer higher weights to recent changes in the data, while the value of θ nearer to zero has higher smoothing effect and is less responsive to recent changes.

There is no precise method of choosing θ. Accurate factors are selected on the basis of the statistician's judgments or else a statistical technique may be used to optimize the value of θ. For example, the method of least squares can be used to estimate the value of θ for which the sum of the quantities is diminished.

Exponential Smoothing Methods

The three exponential smoothing methods to determine exponential smoothing. They are:

Simple or single exponential smoothing

Double exponential smoothing

Triple exponential smoothing

Single Exponential Smoothing

If the data which is obtained has no trend and no seasonal pattern, then the single exponential smoothing method for forecasting the time series is primarily used. This method makes use of weighted moving averages with exponentially decreasing weights.

The single exponential smoothing method formula is given by:

st = θxt+(1 – θ)st-1 = st-1 + θ(xt – st-1)

Double Exponential Smoothing

The double exponential smoothing method is also known as Holt's trend corrected or second-order exponential smoothing. This method is primarily used to forecast the time series when the data has a linear trend and no seasonal pattern. The motive of double exponential smoothing is to introduce a term considering the possibility of a series indicating some form of trend. This slope component is itself reformed through exponential smoothing.

The double exponential smoothing formula is derived by:

S1 = y1

B1 = y1-y0

For t>1,

st = θyt + (1 – θ)(st-1 + bt-1)

βt = β(st – st-1) + (1 – β)bt-1

Here,

St is smoothed statistic, it is the simple weighted average of present observation yt

st-1 = former smoothed statistic

θ = smoothing factor of data; 0 < θ < 1

t = time period

bt = accurate estimation of trend at time t

β = trend smoothing factor; 0 < β <1

Triple Exponential Smoothing

In the triple exponential smoothing method, exponential smoothing is used thrice. This method is primarily used to forecast the time series when the data has both linear trend and seasonal patterns.This method is also known as holt-Winters exponential smoothing.

The triple exponential smoothing formula is derived by:

s\[_{0}\] = x\[_{0}\]

s\[_{t}\] = α\[\frac{x_{t}}{c_{t-L}}\] + (1 - α)(s\[_{t-1}\] + b\[_{t-1}\])

b\[_{t}\] = β(s\[_{t}\] - s\[_{t-1}\] + (1 - β)b\[_{t-1}\]

c\[_{t}\] = γ\[\frac{x_{t}}{s_{t}}\] + (1 - γ)c\[_{t-L}\]

Here,

st = smoothed statistic, it is the simple weighted average of present observation xt

st-1 = previous smoothed statistic

α = smoothing factor of data; 0 < α < 1

t = time period

bt = accurate estimation of trend at time t

β = trend smoothing factor; 0 < β <1

ct = sequence of seasonal error-free factors at time t

γ = seasonal variation smoothing factor; 0 < γ < 1

Solved Examples

1. The Sales of Books in a Bookstall for the Last 10 Months is Given Below in Tabulated Form. Calculate the Simple Exponential Smoothing Estimating α =0.3 for the Below Data.

Quiz Time

1. The Use of Smoothing Technique is Accurate When

The primary source of variation is random behavior

Seasonality is includes

Data exhibit a strong trend

All the above are accurate

2. Times Series Data is Classified in

Two components

Three components

Four components

Five components

FAQ (Frequently Asked Questions)

1. What are the Advantages of Exponential Smoothing?

Advantages of exponential smoothing are:

1. Exponential Smoothing is Easy to Learn and Use - Only three parts of information are needed for exponential smoothing method. The first information which is required is the forecast for the latest time period. Second, it required actual value for that time period. And the last, it requires the value of the smoothing constant, a weighting factor that reflects the weight assigned to the latest data values.

2. It Provides Accurate Forecasts- An exponential smoothing method delivers a forecast of one year ahead.The forecast of further years can be generated using the trend projection. The forecast is considered reliable as it considers the difference between the actual project and what actually appeared.

3. It Gives High Priority to Recent Observation- Observed information is the sum of two or more components, one being the random error which is the difference between the observed value and true value. In exponential smoothing techniques, random changes are ignored. As such it is easier to see the underlying phenomena.

2. What are the Disadvantages of Exponential Smoothing?

Disadvantages of exponential smoothing are:

1. It Delivers Forecasts that Lag Behind the Actual Trend- The lag is the adverse effect of the exponential smoothing process. There’s a reason this method has smoothing in its name as it ignores the up and down connected with random changes. If we observe exponential smoothing on a graph, we can see a smoother line or curve. But not considering the random changes enables us to see the underlying phenomena which help at the time of presenting data and making a forecast of upcoming values.

2. It Cannot Manage Trends Accurately- Exponential smoothing can be precisely used for the forecast that is short terms and in the absence of seasonal and cyclic fluctuations. Due to this, forecasts are not precise when data with cyclic or seasonal changes are present. As such, this kind of averaging does not work accurately if there is a trend in the series. These methods can be primarily used when a reasonable amount of continuity between the present and past can be predicted. As such, it is well-suited for short-term forecasting as it predicts future patterns and trends that look like current trends and patterns. These kinds of observations make sense in the short term, as it creates issues the further the forecast approaches.