MM3425 BUSINESS ANA LYTICS

TIME SERIES ANALYSIS AND FORECASTING

1
FOREC ASTING IS A VITAL FUNCTION

◾ Forecasting is a vital function and affects every significant management decision

◾ Finance and accounting use forecasts as the basis for budgeting and cost control
◾ Marketing relies on forecasts to make key decisions such as new product planning and
personnel compensation
◾ Production uses forecasts to select suppliers; determine capacity requirements; and drive
decisions
about purchasing, staffing, and inventory

2
 FORECASTING

■ Qualitative forecasting methods

– Qualitative methods generally involve the use of expert judgment to develop forecasts
■ Quantitative forecasting methods can be used when
– Past information about the variable being forecast is available
– Information can be quantified
– Reasonable to assume that past is prologue
■ Time series analysis
– The objective of time series analysis is to uncover a pattern in the time series and then
extrapolate the pattern into the future
■ Causal relationship forecasting
– Relating demand to an underlying factor other than time
3
FOREC ASTING TECHNIQUES A N D C O M M O N MODELS

4
FOREC ASTING TECHNIQUES A N D C O M M O N MODELS

5
FOREC ASTING TECHNIQUES A N D C O M M O N MODELS

6
FORECASTING MODEL SELECTION

◾ Time horizon to forecast

◾ Accuracy required
◾ Data availability
◾ Size of forecasting budget
◾ Availability of qualified personnel

7
C O M PARISON OF FOREC ASTING TECHNIQUES

8
TIME SERIES

■ Time series
– A sequence of observations on a variable measured at successive
points in time or over successive periods of time
– The measurements may be taken every hour, day, week, month, year, or
any other regular interval
– If the behavior of the times series data of the past is expected
to continue in the future, it can be used as a guide in selecting
an appropriate forecasting method

9
EUR/CHF

10
 TIME SERIES PATTERN

■ The underlying pattern in the time series is

an important factor in selecting a
forecasting method
■ A time series plot should be one of the first
analytic tools
■ We need to use a forecasting method that is
capable of handling the pattern exhibited by the
time series effectively

11
TIME SERIES PATTERN

■ Horizontal pattern
■ Trend pattern
■ Seasonal pattern
■ Cyclical pattern
■ Random variation

12
TIME SERIES PATTERN

Horizontal pattern
■ Stationary time series
– It denotes a time series whose statistical properties are
independent of time
■ The process generating the data has a constant mean
■ The variability of the time series is constant over time
■ A time series plot for a stationary time series will always exhibit a
horizontal pattern with random fluctuations
13
14
15
C O M M O N TYPES OF TRENDS

16
 SIMPLE MO VING AVERAGE

■ Simple moving average method

– Use the average of the most recent
k data values in the time series as
the forecast for the next period
– Useful when demand is neither
growing nor declining rapidly and
does not have seasonal
characteristics
– Longer gives more smoothing
(i.e., larger k)
– Shorter reacts quicker to trends
17
SIMPLE MO VING AVERAGE EXAMPLE

1000 + 1400 + 800

=
10
67
3

1500 + 1000 + 1400

=
1
3
0
0
3
1500 + 1500 + 1000
=
1
3
3
3
3

18
SIMPLE MO VING AVERAGE EXAMPLE

19
(17+21+19)/3=19

20
WEIGHTED MO VING AVERAGE

◾ The moving average formula implies an equal weight being placed on each value
that is being averaged
◾ The weighted moving average permits an unequal weighting on prior time periods

◾ All the weights must sum to one

◾ w1>w2>wk (e.g., if k=3, there are 3 weights)

ŷt +1 = wkyt-k+1 +...+ w2yt-1 +w1yt

= w1yt + w2yt-1+...+wkyt-k+1 21
WEIGHTED MO VING AVERAGE EXAMPLE

3 Week 3 Week
Week Demand 3 Week W1=0.7, W2=0.2, W3=0.1 W1=0.5, W2=0.3, W3=0.2

1 800
2 1400
3 1000
4 1500 1067 1060 1080
5 1500 1300 1390 1330
6 1300 1333 1450 1400
7 1800 1433 1360 1400
8 1700 1533 1670 1590
9 1300 1600 1680 1650
10 1700 1600 1430 1520
11 1700 1567 1620 1580
12 1500 1567 1660 1620
13 2300 1633 1560 1600
22
14 2300 1833 2080 1940
15 2000 2033 2220 2140
WEIGHTED MO VING AVERAGE EXAMPLE
Actual 3-Week 3-W eek (W eights: 0.7, 0.2, 3-W eek (W eights: 0.5, 0.3,
D emand (Simple) 0.1) 0.2)
2500

200
0

150
0
Demand

100
0

500

0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 23
Week
CHOOSING WEIGHTS

◾ Experience and trial-and-error are the simplest

ways
◾ Generally, the most recent past is the best
indicator

24
EXPONENTIAL SMOOTHING

■ Exponential smoothing
– Use a weighted average of past time series values as a
forecast

= ŷt + α(yt-ŷt)

■ Smoothing
– Weight constant (α) the actual value in period t; weight given to the forecast
given to
in period t is 1-α
– α is between 0 and 1
– Use small values if demand is stable, larger values if demand is
fluctuating 25

– Choose α that minimizes the error

FIRST FORECAST

◾ No way to find ŷ1 since ŷt is a function of ŷt-1

◾ Estimate
◾ The first period’s demand
◾ Using an average of preceding periods, such as the average of the first two or three
periods

26
Damping factor = 1- α How to make them integers?
27
EXPONENTIAL SMOOTHING EXAMPLE
Actual D emand Exponential Smoothing (Alpha=0.2)
Week Demand α=0.2 α=0.6 Exponential Smoothing (Alpha=0.6)
1 800 800 800 2500
2 1400 800 800
3 1000 920 1160 2000
4 1500 936 1064
5 1500 1049 1326 1500

Demand
6 1300 1139 1430
7 1800 1171 1352 1000

8 1700 1297 1621

500
9 1300 1378 1668
10 1700 1362 1447
0
11 1700 1430 1599 1
12 1500 1484 1660
2
13 2300 1487 1564 ŷ4= ŷ3+ɑ(y
3 3
- ŷ3)=920+0.2*(1000-920)=936
14 2300 1650 2006
ŷ4= ŷ3+ɑ(y4 3- ŷ3)=1160+0.6*(1000-1160)=1064
28
15 2000 1780 2182
5
EXPONENTIAL SMOOTHING
◾ A weighted average method that includes all past data in the forecasting
calculation
◾ More recent results weighted more heavily
◾ Does not require large amounts of historical data
◾ Widely accepted because
◾ Exponential models are surprisingly accurate
◾ Formulating an exponential model is relatively easy
◾ The user can understand how the model works
◾ Little computation is required to use the model
◾ Limited use of historical data

29
TREND ADJUSTED EXPONENTIAL SMOOTHING

ŷt +1 = αyt + (1– α)FITt or FITt + α(yt – FITt)

Tt +1 = Tt + δ(ŷt +1 – FITt ) or Tt +1 = Tt + αδ(yt – FITt)
FITt +1 = ŷt +1 + Tt +1
ŷt +1 = Expoentially smoothed forecast for period t+1
Tt +1 = Expoentially smoothed trend for period t+1
FITt +1= Forecast including trend for period t+1
FITt = Forecast including trend for period t
yt = Actual demand for period t
30
α = Alpha smoothing constant
TREN D ADJUSTED EXPONENTIAL SMOOTHING

Demand t=1
Week (yt) ŷt Tt FITt
1 800 700 100 800 ŷ2 = FIT1 + α(y 1 – FIT 1)=800+0.2*(800-800)=800
2 1400 800 100 900
3 1000 1000 130 1130
T2 = T1 + δ(ŷ2 – FIT1) =100+0.3*(800-800)=100
4 1500 FIT2= ŷ2+ T2=800+100=900
5 1500
6 1300
7 1800
8 1700 t=2
9 1300 ŷ3 = FIT 2 + α(y 2 – FIT 2)=900+0.2*(1400-900)=1000
10 1700 T3 = T2 + δ(ŷ3 – FIT2)=100+0.3*(1000-900)=130
11 1700
12 1500 FIT3= ŷ3+ T3=1000+130=1130
13 2300
14 2300
15 2000
α=0.2, δ=0.3 31
TREN D ADJUSTED EXPONENTIAL SMOOTHING

Demand
Week (yt) ŷt Tt FITt
1 800 700 100 800
Acutal Trend Adjusted Exponential
2 1400 800 100 900 D emand Smoothing
2500
3 1000 1000 130 1130
4 1500 1104 122.2 1226 2000
5 1500 1280.8 138.64 1419
6 1300 1435.2 143.5 1579 1500

Demand
7 1800 1523.2 126.76 1650
8 1700 1680 135.76 1816 1000
9 1300 1792.8 128.8 1922
10 1700 1797.6 91.48 1889 500
11 1700 1851.2 80.14 1931
12 1500 1884.8 66.28 1951 0
1 2 3 4 5 6 7 8 9 10 11 12 13 14
13 2300 1860.8 39.22 1900 Week 15
14 2300 1980 63.22 2043
15 2000 2094.4 78.64 2173
32
α=0.2, δ=0.3