2. INTRODUCTION
• The idea of Holt Winters exponential smoothing is to smooth the
original univariate series ( series that consists of single (scalar) observations
recorded sequentially over equal time increments.)and to use the smoothed
series in forecasting future values of the variable of interest.
• Exponential Smoothing assigns exponentially decreasing weights as
the observation get older. In other words, recent observations are
given relatively more weight in forecasting than the older
observations.
• Holt Winters Algorithms are applied when data is stationary
3. THERE ARE THREE TYPES OF HOLT WINTERS EXPONENTIAL
SMOOTHING
• Holt Winters Single Exponential Smoothing : Suitable for
forecasting data with no trend or seasonal
pattern. However level of the data may be
changing over time as shown in figure 1
• Holt Winters Double Exponential Smoothing : Suitable
for forecasting data with trend as shown in figure
• Holt Winters Triple Exponential Smoothing : Suitable for
forecasting data with trend and/or seasonality as
shown in figure 2.
LEVEL
Level 1
Level 2
Level 3
Figure 1
TREND
SEASONAL
Figure 2
5. • What Is Stationarity?
A stationary time series is one whose statistical properties such as mean, variance are all constant over time.
• What Is Level?
In a time series there might be one or more significant jump, occurring in few successive time points. Post the jump the level
of a time series is different compared to the one before jump. If there is no significant jump, we say the time series has level 1,
else more than one depending on the number of considerable jump(s).
• What is Additive & Multiplicative Time Series
• In case of additive, the observed time series is considered to be the sum of components: seasonal, trend, cyclical and
random
• Whereas in case of multiplicative, the series is multiplication of these components : seasonal, trend, cyclical and random..
6. HOLT WINTERS SINGLE EXPONENTIAL SMOOTHING
• This method is suitable for forecasting when data is stationary and there is only level present in data with no
trend or seasonal pattern, but the level of the time series is slowly changing over time
Limitations
• Holt Winters Exponential Smoothing Algorithms are used only when data series is stationary
• Holt Winters Single Exponential Smoothing is used for forecasting data with only levels and in real
life data this is hardly the case
• Holt winter is only for univariate data forecasts
•Use case
• Forecasting number of viewers by day for a particular game show for next two months
• Input data: Last six months daily viewer count data
• Data pattern : Data taken as an input exhibit stationarity and no trend /seasonality • Business
benefit:
• Helps in planning for repeat telecast
• Can pitch in for more advertisement (fund raise) if projected count of viewers are high
• Improvement planning can be done for the game show to increase/maintain the level of popularity
7. HOLT WINTERS DOUBLE EXPONENTIAL SMOOTHING
• Double exponential smoothing is a single exponential smoothing applied twice to both level and trend and
hence is suitable for forecasting when data is stationary and there is level and trend pattern present in data .
•Limitations
• Holt Winters Double Exponential Smoothing is used for forecasting data consisting only level and
trends . In case of seasonality this algorithm is not applicable
• Holt winter double exponential smoothing is only for univariate and stationary data forecasts
• use case
• Insurance claim manager wants to forecast policy sales for next month based on past 12 months
data
• Data pattern : Input data exhibits stationarity , level and strong upward trend but no seasonality
• Business benefit:
• If projected claims are lower than expected then proper marketing strategy can be devised to
improve sales
• Competition policy can be analyzed in terms of what all perks and benefits they provide to
customers and existing policy can be modified to increase the market share
8. HOLT WINTERS TRIPLE EXPONENTIAL SMOOTHING
• Triple exponential smoothing handles seasonality as well as trend. Seasonality is defined as the
tendency of time-series data to exhibit behavior that repeats itself at regular period of time
• Limitations
• 1.Holt Winters triple exponential smoothing can not handle irregular pattern well.
• 2.Holt winter triple exponential smoothing is only for univariate and stationary data forecasts.
• use case
• A power generator company wants to predict the electricity demand for next two months based on
past 2 years’ daily power consumption data
• Data pattern : Input data exhibits stationarity , trend and seasonality
• Business benefit:
• A power generator company can make use of these forecasts for the control and scheduling of
power systems
• It widely helps in balancing supply and demand