TIME SERIES

MODELS
 Definitions

• Forecast is a prediction of future events

used for planning process.
• Time Series is the repeated observations
of demand for a service or product in their
order of occurrence.
Components of a Time Series
• A time series can consist of five components.
– Horizontal or Stationary – Fluctuations around a constant
mean.
– Long - term trend (T).

Quantity
– Cyclical effect (C).
– Seasonal effect (S).
– Random variation (R).
Time

A trend is a long term relatively

smooth pattern or direction, that
persists usually for more than one year.
A cycle is a wavelike pattern describing
a long term behavior (for more than one
year).

Cycles are seldom regular, and

often appear in combination with
other components.
6/90 6/93 6/96 6/99 6/02

The seasonal component of the time series

exhibits a short term (less than one year)
calendar repetitive behavior.
6/97 12/97 6/98 12/98 6/99
Random variation comprises the irregular
unpredictable changes in the time series.
It tends to hide the other (more
predictable)
components.

Random or Irregular Variation –classified

into:

•Episodic – unpredictable but identifiable

•Residual – also called chance fluctuation
and unidentifiable
 Moving Average and Exponential
Smoothing Methods
• Consider models applicable to time series data with
seasonal, trend, or both seasonal and trend component and
stationary data.
• Forecasting methods discussed can be classified as:
– Averaging methods.
• Equally weighted observations
– Exponential Smoothing methods.
• Unequal set of weights to past data, where the weights decay
exponentially from the most recent to the most distant data points.
• All methods in this group require that certain parameters to be defined.
• These parameters (with values between 0 and 1) will determine the
unequal weights to be applied to past data.
 Moving Averages

– A n-period moving average for time period t is

the arithmetic average of the time series values for
the n most recent time periods.

– For example: A 3-period moving average at period

(t+1) is calculated by (yt-2 + yt-1 + yt)/3
 Centered Moving Average
Method
The centered moving average method
consists of computing an average of n
periods' data and associating it with the
midpoint of the periods. For example, the
average for periods 5, 6, and 7 is
associated with period 6.
 Moving Averages
• A large k is desirable when there are wide,
infrequent fluctuations in the series.
• A small k is most desirable when there are
sudden shifts in the level of series.
• For quarterly data, a four-quarter moving
average, MA(4), eliminates or averages out
seasonal effects.
 Moving Averages
• For monthly data, a 12-month moving
average, MA(12), eliminate or averages out
seasonal effect.
• Equal weights are assigned to each observation
used in the average.
• Each new data point is included in the average
as it becomes available, and the oldest data
point is discarded.
 Summary of Moving Averages
• Advantages of Moving Average Method
– Easily understood
– Easily computed
– Provides stable forecasts

• Disadvantages of Moving Average Method

– Requires saving all past n data points
– Lags behind a trend
– Ignores complex relationships in data
 Simple Moving Average – Example (File: PPT_MA)
Period Actual MA3 MA5

1 42    

• Consider the following 2 40    

data,
3 43    

• Starting from 4th period

4 40 41.7  
one can start forecasting
by using MA3. Same is 5 41 41.0  

true for MA5 after the 6th 6 39 41.3 41.2

period. 7 46 40.0 40.6

• Actual versus 8 44 42.0 41.8

predicted(forecasted) 9 45 43.0 42

graphs are as follows; 10 38 45.0 43

11 40 42.3 42.4

12   41.0 42.6
 Simple Moving Average -
Example
Actual
MA5
47
45
43
41
39
37 MA3
35
1 2 3 4 5 6 7 8 9 10 11 12
 Sample Excel analysis for Actual,
Moving Averages for 3 and 5 periods
Period Actual MA3 MA5
1 42 #N/A #N/A
2 40 #N/A #N/A
3 43 41.66667 #N/A
4 40 41 #N/A
5 41 41.33333 41.2
6 39 40 40.6
7 46 42 41.8
8 44 43 42
9 45 45 43
10 38 42.33333 42.4
11 40 41 42.6
 Sample Excel analysis for Actual, Centred
Moving Averages for 2 and 4 periods
Period Actual Avg2 Centered MA2 Avg4 Centered MA4
1 42 NA NA
41
2 40 41.25 NA
41.5 41.25
3 43 41.5 41.125
41.5 41
4 40 41 40.875
40.5 40.75
5 41 40.25 41.125
40 41.5
6 39 41.25 42
42.5 42.5
7 46 43.75 43
45 43.5
8 44 44.75 43.375
44.5 43.25
9 45 43 42.5
41.5 41.75
10 38 40.25 NA
39
11 40 NA NA
Sample SPSS analysis for Actual,
Centered MA3 and Prior MA3

Period Actual MA3 Prior_MA3

1 42

2 40 41.7

3 43 41.0

4 40 41.3 41.7

5 41 40.0 41.0

6 39 42.0 41.3

7 46 43.0 40.0
Sample SPSS graph for Actual,
Centered MA5 and Prior MA5
Example: Weekly Department Store Sales
Period (t) Sales (y)
• The weekly sales 1
2
5.3
4.4

figures (in millions of 3

4
5.4
5.8
5 5.6
dollars) presented in 6
7
4.8
5.6

the following table are 8

9
5.6
5.4
10 6.5
used by a major 11
12
5.1
5.8

department store to 13
14
5
6.2

determine the need for

15 5.6
16 6.7
17 5.2
temporary sales 18
19
5.5
5.8

personnel. 20
21
5.1
5.8
22 6.7
23 5.2
24 6
25 5.8
 Example: Weekly Department Store Sales

• Use a three-week moving average (k=3) for

the department store sales to forecast for the
week 24 and 26.
( y23  y22  y21 ) 5.2  6.7  5.8
yˆ 24    5.9
3 3

The forecast for the week 26 is

y25  y24  y23 5.8  6  5.2
yˆ 26    5.7
3 3
 Example: Weekly Department Store Sales
Period (t) Sales (y) forecast
Weekly Sales Forecasts 1 5.3
2 4.4
8 3 5.4
4 5.8 5.033333
5 5.6 5.2
7
6 4.8 5.6
7 5.6 5.4
6 8 5.6 5.333333
9 5.4 5.333333
10 6.5 5.533333
5
11 5.1 5.833333
12 5.8 5.666667
Sales (y)
4 13 5 5.8
s
le

forecast
a
S

14 6.2 5.3
15 5.6 5.666667
3
16 6.7 5.6
17 5.2 6.166667
2 18 5.5 5.833333
19 5.8 5.8
20 5.1 5.5
1
21 5.8 5.466667
22 6.7 5.566667
0 23 5.2 5.866667
0 5 10 15 20 25 30 24 6 5.9
Weeks 25 5.8 5.966667
5.666667
 Problem: Robert’s Drugs
During the past ten weeks, sales of cases of Comfort
brand headache medicine at Robert's Drugs have been as
follows:
Week Sales Week Sales
1 110 6 120
2 115 7 130
3 125 8 115
4 120 9 110
5 125 10 130
If Robert's uses a 3-period moving average to forecast
sales, what is the forecast for Week 11?
 Exponential Smoothing Methods
• This method provides an exponentially
weighted moving average of all previously
observed values.
• Appropriate for data with no predictable
upward or downward trend.
• The aim is to estimate the current level and use
it as a forecast of future value.
 Simple Exponential Smoothing Method

• Formally, the exponential smoothing equation is

Ft 1   yt  (1   ) Ft
• Ft 1  forecast for the next period.
  = smoothing constant.
• yt = observed value of series in period t.
• Ft = old forecast for period t.
– The forecast Ft+1 is based on weighting the most recent
observation yt with a weight  and weighting the most
recent forecast Ft with a weight of 1- 
 Simple Exponential Smoothing Method

• The implication of exponential smoothing can

be better seen if the previous equation is
expanded by replacing Ft with its components
as follows:

Ft 1   yt  (1   ) Ft
  yt  (1   )[ yt 1  (1   ) Ft 1 ]
  yt   (1   ) y t 1 (1   ) 2 Ft 1
 Simple Exponential Smoothing Method

• If this substitution process is repeated by

replacing Ft-1 by its components, Ft-2 by its
components, and so on the result is:
Ft 1   yt   (1   ) y t 1  (1   ) 2 y t  2   (1   )3 y t 3    (1   )t 1 y1

• Therefore, Ft+1 is the weighted moving average

of all past observations.
 Simple Exponential Smoothing Method

• The exponential smoothing equation rewritten

in the following form elucidate the role of
weighting factor .
Ft 1  Ft   ( yt  Ft )

• Exponential smoothing forecast is the old

forecast plus an adjustment for the error that
occurred in the last forecast.
 Simple Exponential Smoothing Method
• The value of smoothing constant  must be between
0 and 1
  can not be equal to 0 or 1.
• If stable predictions with smoothed random variation
is desired then a small value of  is desire.
• If a rapid response to a real change in the pattern of
observations is desired, a large value of  is
appropriate.
 Measures of Forecast Errors
• Cumulative Forecast Error (CFE)
• Mean Squared Error (MSE)
• Standard Deviation (σ)
• Mean Absolute Deviation (MAD)
• Mean Absolute Percent Error (MAPE)
Choosing a Method
Forecast Error

Measures of Forecast Error

Et = yt – Ft

CFE = Et
(Et – E )2
 = Et2 n–1
MSE =
n
|Et | [ |Et | (100) ] / yt
MAD = MAPE = n
n
Example
The following table shows the actual
sales of upholstered chairs for a
furniture manufacturer and the
forecasts made for each of the last
eight months. Calculate CFE, MSE, σ,
MAD and MAPE for this product.
Choosing a Method
Forecast Error

Absolute
Error Absolute Percent
Month, Demand, Forecast, Error, Squared, Error, Error,
t yt Ft Et Et2
|Et| (|Et|/yt)(100)
1 200 225 -25 625 25 12.5%
2 240 220 20 400 20 8.3
3 300 285 15 225 15 5.0
4 270 290 –20 400 20 7.4
5 230 250 –20 400 20 8.7
6 260 240 20 400 20 7.7
7 210 250 –40 1600 40 19.0
8 275 240 35 1225 35 12.7
Total –15 5275 195 81.3%
Choosing a Method
Forecast Error
Measures of Error
Absolute
CFE = – 15 Error Absolute Percent
Month, Demand, Forecast, Error, Squared, Error, Error,
– 15 D
E =t = –t 1.875 Ft Et Et2
|Et| (|Et|/Dt)(100)
8
1 200 225 –25 625 25 12.5%
2 240
5275 220 20 400 20 8.3
MSE 3= 300 = 659.4
285 15 225 15 5.0
4 8 270 290 –20 400 20 7.4
5 230 250 –20 400 20 8.7
6  = 27.4
260 240 20 400 20 7.7
7 210 250 –40 1600 40 19.0
8 195
275= 24.4 240 35 1225 35 12.7
MAD =
8 Total –15 5275 195 81.3%

81.3%
MAPE = = 10.2%
8
Measures of Forecast Accuracy
• Choose MSE if it is important to avoid (even a
few) large errors. Otherwise, use MAD / MAPE.

• A useful procedure for model selection.

– Use some of the observations to develop
several competing forecasting models.
– Run the models on the rest of the
observations.
– Calculate the accuracy of each model.
– Select the model with the best accuracy
measure.
 Example: Robert’s Drugs
During the past ten weeks, sales of cases of Comfort
brand headache medicine at Robert's Drugs have been
as follows:
Week Sales Week Sales
1 110 6 120
2 115 7 130
3 125 8 115
4 120 9 110
5 125 10 130
If Robert's uses exponential smoothing to forecast
sales, which value for the smoothing constant ,  = .1
or  = .8, gives better forecasts?
 Example: Robert’s Drugs

• Exponential Smoothing
To evaluate the two smoothing constants,
determine how the forecasted values would
compare with the actual historical values in each
case.
Let: Yt = actual sales in week t
Ft = forecasted sales in week t
F1 = Y1 = 110
For other weeks, Ft+1 = .1Yt + .9Ft
 Example: Robert’s Drugs
• Exponential Smoothing
For  = .1, 1 -  = .9
F1 = 110
F2 = .1Y1 + .9F1 = .1(110) + .9(110) = 110
F3 = .1Y2 + .9F2 = .1(115) + .9(110) = 110.5
F4 = .1Y3 + .9F3 = .1(125) + .9(110.5) = 111.95
F5 = .1Y4 + .9F4 = .1(120) + .9(111.95) = 112.76
F6 = .1Y5 + .9F5 = .1(125) + .9(112.76) = 113.98
F7 = .1Y6 + .9F6 = .1(120) + .9(113.98) = 114.58
F8 = .1Y7 + .9F7 = .1(130) + .9(114.58) = 116.12
F9 = .1Y8 + .9F8 = .1(115) + .9(116.12) = 116.01
F10= .1Y9 + .9F9 = .1(110) + .9(116.01) = 115.41
 Example: Robert’s Drugs
• Exponential Smoothing
For  = .8, 1 -  = .2
F1 = 110
F2 = .8(110) + .2(110) = 110
F3 = .8(115) + .2(110) = 114
F4 = .8(125) + .2(114) = 122.80
F5 = .8(120) + .2(122.80) = 120.56
F6 = .8(125) + .2(120.56) = 124.11
F7 = .8(120) + .2(124.11) = 120.82
F8 = .8(130) + .2(120.82) = 128.16
F9 = .8(115) + .2(128.16) = 117.63
F10= .8(110) + .2(117.63) = 111.53
 Example: Robert’s Drugs
• Mean Squared Error
In order to determine which smoothing
constant gives the better performance,
calculate, for each, the mean squared
error for the nine weeks of forecasts,
weeks 2 through 10 by:
[(Y2-F2)2 + (Y3-F3)2 + (Y4-F4)2 + . . . +
(Y10-F10)2]/9
Example: Robert’s Drugs
 = .1  = .8
Week Yt Ft (Yt - Ft)2 Ft (Yt - Ft)2
1 110
2 115 110.00 25.00 110.00 25.00
3 125 110.50 210.25 114.00 121.00
4 120 111.95 64.80 122.80 7.84
5 125 112.76 149.94 120.56 19.71
6 120 113.98 36.25 124.11 16.91
7 130 114.58 237.73 120.82 84.23
8 115 116.12 1.26 128.16 173.30
9 110 116.01 36.12 117.63 58.26
10 130 115.41 212.87 111.53 341.27
Sum 974.22 Sum 847.52
MSE Sum/9 108.25
108.25
Sum/9 94.17
94.17
 Seasonal Analysis
• Seasonal variation may occur within a
year or within a shorter period (month,
week)
• To measure the seasonal effects we
construct seasonal indices.
• Seasonal indexes express the degree to
which the seasons differ from the average
time series value across all seasons.
 Computing Seasonal Indices
• Remove the effects of the seasonal and
random variations by regression analysis
ŷt = b0 + b1t
• For each time period compute the ratio
yt/yt > This is based on the Multiplicative Model.

which removes most of the trend variation

>
• For each season calculate the average of yt/yt
which provides the measure of seasonality.
• Adjust the average above so that the sum of averages
of all seasons is 1 (if necessary)
 Computing Seasonal Indices
• Example
– Calculate the quarterly seasonal indices for
hotel occupancy rate in order to measure
seasonal variation.
– Data:
Year Quarter Rate Year Quarter Rate Year Quarter Rate
1996 1 0.561 1998 1 0.594 2000 1 0.665
2 0.702 2 0.738 2 0.835
3 0.8 3 0.729 3 0.873
4 0.568 4 0.6 4 0.67
1997 1 0.575 1999 1 0.622
2 0.738 2 0.708
3 0.868 3 0.806
4 0.605 4 0.632
 Computing Seasonal Indices
• Perform regression analysis for the model
y = 0 + 1t +  where t represents the time,
and y represents the occupancy rate.

ŷ  .639368  .005246t
Time (t) Rate
1 0.561
2 0.702
3 0.800
Rate

4 0.568
5 0.575
6 0.738
7 0.868 0 5 10 15 20 25
8 0.605 The regression linet represents trend.
. .
. .
 The Ratios

>
yt  y t
t yt ŷ t Ratio
1 .561 .645 .561/.645=.870
2 .702 .650 .702/.650=1.08
3 ………………………………………………….
=.639368+.005245(1)
No trend is observed, but
yt seasonality and randomness
Rate/Predicted rate
ŷ t still exist.
1.5

0.5

0
1

11

15

17
13

19
 The Average Ratios by Seasons
Rate/Predictedrate
Rate/Predicted rate
0.870
0.870
1.080
1.080
• To remove most of the random variation
1.221
1.221 but leave the seasonal effects,average
0.860
0.860
0.864
0.864
the terms yt / yˆ t for each season.
1.100
1.100
1.284
1.284 Rate/Predicted rate
0.888
0.888
0.865
0.865 1.5
1.067
1.067
1.046
1.046 1
0.854
0.854
0.879 0.5
0.879
0.993
0.993 0
1.122
1.122
0.874 1 3 5 7 9 11 13 15 17 19
0.874 Average ratio for quarter 1: (.870 + .864 + .865 + .879 + .913)/5 = .878
0.913
0.913
1.138
1.138
Average ratio for quarter 2: (1.080+1.100+1.067+.993+1.138)/5 = 1.076
1.181
1.181 Average ratio for quarter 3: (1.221+1.284+1.046+1.122+1.181)/5 = 1.171
0.900
0.900
Average ratio for quarter 4: (.860 +.888 + .854 + .874 + .900)/ 5 = .875
 Adjusting the Average Ratios
• In this example the sum of all the averaged ratios
must be 4, such that the average ratio per season is
equal to 1.
• If the sum of all the ratios is not 4, we need to adjust
them proportionately.
Suppose the sum of ratios is equal to 4.1. Then each
ratio will be multiplied by 4/4.1.

(Seasonal averaged ratio) (number of seasons)

Seasonal index =
Sum of averaged ratios

In our problem the sum of all the averaged ratios is equal to 4:

.878 + 1.076 + 1.171 + .875 = 4.0.
No normalization is needed. These ratios become the
seasonal indices.
 Interpreting the Seasonal Indices
• The seasonal indexes tell us what is the ratio
between the time series value at a certain
season, and the overall seasonal average.
17.1% above the
• In our problem: annual average
117.1%
7.6% above the 12.5% below the
annual average annual average
Annual average
occupancy (100%) 107.6%
87.8% 87.5%
12.2% below the
annual average

Quarter 1 Quarter 2 Quarter 3 Quarter 4 Quarter 1 Quarter 2 Quarter 3 Quarter 4

 The Smoothed Time Series
• The trend component and the seasonality
component are recomposed using the
multiplicative model.

ŷ t  T̂t  Ŝ t  (.639  .0052t )Ŝ t

In period #1 ( quarter 1): ŷ1  T̂1  Ŝ1  (.639  .0052(1))(.878)  .566

In period #2 ( quarter 2): ŷ 2  T̂2  Ŝ 2  (.639  .0052(2))(1.076)  .699
0.9
Actual series Smoothed series
0.8

0.7

0.6

0.5 The linear trend (regression) line

1 3 5 7 9 11 13 15 17 19
 Deseasonalized Time Series

Seasonally adjusted time series = Actual time series

Seasonal index

By removing the seasonality, we can

identify changes in the other components of
the time series, that might have occurred
over time.
 Deseasonalized Time Series

In period #1 ( quarter 1): y1 / SI1  .561 / .878  .639

In period #4 ( quarter 2): y 2 / SI2  .708 1.076  .652
In period #5 ( quarter 1): y5 / SI1  .575 .878  .661
There was a gradual increase in occupancy rate
1
0.8

0.6
0.4
0.2
0
0 5 10 15 20 25