Professional Documents
Culture Documents
1 2 3
Resource Sales and operations Demand
Manufacturing Planning & Control (1-9) planning planning management
Enterprise Resource Planning (1-9) 4
Demand Management (3) Master production
scheduling Front End
Forecasting (3)
Sales & Operations Planning (1-5)
5 6
Master Production Scheduling (4) Detailed capacity Detailed material
Capacity Planning & Management (6-7) planning planning
Production Activity Control (5-9) Engine
7
Advanced Scheduling (8) Material and
Just-In-Time (1-9) capacity plans
Distribution Requirements (6, 9)
Management of Supply Chain Logistics (6,9) 8 9
Shop-floor Supplier Back End
Order Point Inventory Control Methods (3,6)
systems systems
(Demand) Forecasting
Process of predicting a future event
Underlying basis of all business decisions:
Production
Inventory
Personnel
Facilities
Qualitative Forecasting Approaches
Jury of Executive Opinion
Delphi Method
Sales Force Composite
Consumer Market Survey
Trend Projection
Naïve Approach
Quantitative Forecasting Methods Averages Methods
Time Series Exponential Smoothing
Associative Models
Linear Regression
• Forecast Accuracy Evaluation Correlation Analysis
3-3
Forecasts common features:
3-4
Quantitative Approaches
TREND PROJECTIONS (PATTERNS)
• based on sequence of evenly spaced data points
• models that base predictions on historical patterns of numerical
data.
• Data can be represented via a TREND PATTERN, SEASONALITY
PATTERN, CYCLE PATTERN, AUTOCORRELATION, RANDOM
VARIATION
?
Non-Linear Trend
TREND PATTERNS
• Develop an equation that describes the trend
• Look at historical data
• Linear vs Non-Linear Trends
Linear Trend
SEASONALITY PATTERN
• models that base predictions on historical patterns of numerical data but
incorporate amplitude of seasonal variation
• Last less than one year
Forecasting with Seasonality
The components of a time series:
Time Series
yt Tt S t Rt
where:
yt = The time series value at period t
Tt = The trend component at period t
St = The seasonal component at period t
Rt = The random component at period t
Multiplicative Decomposition
of a Time Series
Use these steps for the multiplicative decomposition
technique for a forecast:
Centered Ratio-to-
Centered
Quarter Sales Moving Moving-
Period
Average Average
3 2011-Q3 108 123.375 0.8754
4 2011-Q4 135 129.000 1.0465 Example:
4
NSF1 ( SF1 ) (1.0076)(0 .8955) 0.8958
3.9699
SF1 = 0.8890
4
SF2 = 1.1320 NSF2 ( SF2 ) (1.0076)(1 .1320) 1.1406
3.9699
SF3 = 0.8865 4
NSF3 ( SF3 ) (1.0076)(0 .8865) 0.8932
SF4 = 1.0624 3.9699
Total = 3.9699 4
NSF4 ( SF4 ) (1.0076)(1 .0624) 1.0704
3.9699
Total = 4.0000
Multiplicative Decomposition
of a Time Series
yt
Tt Rt
St
Multiplicative Decomposition
of a Time Series
Example:
135
126.1
1.0704
Multiplicative Decomposition
of a Time Series
Tt 116.8545 2.522 t
Multiplicative Decomposition
of a Time Series
4. Generate Forecasts with the Trend and
Seasonal Components
• Formula for Forecasting with the Multiplicative
Decomposition Method
Ft Tt St
Multiplicative Decomposition
of a Time Series
Period of 2 periods
time since ago
‘Last
Season’
Stable Time Series
n A F
1 10 Example
2 10
3 10 What is the forecast for Period 9?
4 10
5 10
6 10 F9 = A9-8 or F9 = A8 or F9 =
7 10 10
8 10
9
Seasonal Variation
Season n A F Example
Spring 1 12 What is the forecast for Periods 9, 10, 11, 12 ?
Summer 2 25
Fall 3 8
Winter 4 2 F9 = A9-4 or F9 = A5 or F9 = 10
Spring 5 10
Summer 6 28
Fall 7 7 F10 = A10-4 or F10 = A6 or
Winter 8 0
Spring 9 F10 = 28
Summer 10
Fall 11
Winter 12 F11 = A11-4 or F11 = A7 or F11 = 7
or
F17 = 34
Averaging Methods 1. Moving average
• Smoothing of random variation 2. Weighted moving
• 3 techniques: average
3. Exponential smoothing
1
Moving
: Ft
Demand in previous n periods
Average n
2
Exponentia l
3 : Ft Ft 1 At 1 Ft 1
Smoothing Relevance
%
F = forecast A = actual = smoothingconstant
Moving
: Ft
Demand in previous n periods
Moving Average Average n
n A F
1 2
Example 1
2 4 What is the 4-Period Moving Average forecast
3 6
4 8
for Period 17?
5 10
6
7
12
14
F17 = Sum (32+30+28+26) / 4
8 16
9
10
18
20
or
11 22
12 24 F17 = 29
13 26
14 28
15 30 Example 2
16 32
17 What is the 6-Period Moving Average forecast
for Period 17?
or
F17 = 27
Weighted Moving
Average
n A Weight % F Example 1
1 2 0
2 4 0
What is the 4-period weighted moving average
3 6 0 forecast for period 17 using a weight of 0.4 for
4
5
8
10
0
0
the most recent period, 0.3 for the next, 0.2 for
6 12 0 the next, and 0.1 for the next.
7 14 0
8 16 0
9 18 0
10 20 0 F17 = Sum (32x0.4) + (30x0.3) + (28*0.2) +
11 22 0
12 24 0 (26*0.1)
13 26 10
14 28 20
15 30 30 Or
16 32 40
17
F17 = 30
Exponentia l
: Ft Ft 1 At 1 Ft 1
Exponential Smoothing
Smoothing
n Actual Forecast
1 65 60
2 55
3 58
4 64
5 Example 1
What is the Period 5 forecast using
smoothing constant of 0.40 ?
or
F5 = 60.83
Exponentia l
: Ft Ft 1 At 1 Ft 1
Exponential Smoothing
Smoothing
n Actual Forecast
1 65
2 55
3 58
4 64
5 Example 2
What is the Period 5 forecast using
smoothing constant of 0.40 ?
(**If no forecast is given – start using Naïve
Method)
or
F5 = 61.48
Equation : yt a bt
n ty t y
n t y
1 2 40
2 4 36
Slope : b 3 6 46
n t t
4 8 32
2 2 5 10 57
6 12 38
7 14 34
y b t
8 16 52
9 18 26
10 20 34
y - intercept : a 11
12
22
24
27
40
n 13
14
15
26
28
30
32
36
37
16 32 32
17 34 26
18 36 36
19 38 32
n ty t y
3 6 46 276 36
4 8 32 256 64
5 10 57 570 100
n t t
7 14 34 476 196
2 2 8 16 52 832 256
9 18 26 468 324
10 20 34 680 400
11 22 27 594 484
y b t
12 24 40 960 576
13 26 32 832 676
y - intercept : a
14 28 36 1008 784
15 30 37 1110 900
n 16
17
18
32
34
36
32
26
36
1024
884
1296
1024
1156
1296
19 38 32 1216 1444
20 40 28 1120 1600
21 42 26 1092 1764
y = a + bt 22
23
44
46
30
28
1320
1288
1936
2116
24 48 32 1536 2304
25 50 40 2000 2500
26 52 32 1664 2704
b = 30(30938) – 930(1045) 27 54 36 1944 2916
28 56 32 1792 3136
30(37820) – (930)(930) = -0.16207 29 58 30 1740 3364
30 60 38 2280 3600
SUM 930 1045 30938 37820
50
40
30
20
10
0
0 10 20 30 40 50 60 70
CORRELATION
• How strong is the linear relationship between the variables?
• Correlation does not necessarily imply causality!
• Coefficient of correlation, r, measures degree of association
• Values range from −1 to +1
n xy x y
r
[n x 2 ( x)2 ][n y 2 ( y )2 ]
Accuracy and Control of Forecasts
Four measures of forecasts are used:
Absolute
Example
What is the MAD of this forecast?
MAD = 10 / 4 or 2.5
Mean Squared Errors (MSE)
Example
What is the MSE of this forecast?
MAD = 30 / 4 or 7.5
Mean Absolute Percent Errors (MAPE)
Example
What is the MAPE of this forecast?
Example
Is this forecast within control limits?
A F A- F
Month (Sales) (Forecast) Error e2
1 90 100 -10 100
2 95 100 -5 25
3 115 100 +15 225
4 100 110 -10 100
5 125 110 +15 225
6 140 110 +30 900
1575
1575
s 262.5 16.2
6
t y ty t squared
1 4200 4200 1
2 4300 8600 4
3 4000 12000 9
4 4400 17600 16
5 5000 25000 25
6 4700 28200 36
7 5300 37100 49
8 4900 39200 64
9 5400 48600 81
10 5700 57000 100
11 6300 69300 121
12 6000 72000 144
330000 A 3766.666667
1716
B 192.30769
Y = 3766 + 192.30t
3-49
3-50
3-51
b. Calculate MSE, MAPE for the model.
11/18/2023 3-52
3-53
MSE
Month Demand Technique 1 Error Error Squared Technique 2 Error Error Squared
1 492 488 4 16 495 -3 9
2 470 484 -14 196 482 -12 144
3 485 480 5 25 478 7 49
4 493 490 3 9 488 5 25
5 498 497 1 1 492 6 36
6 492 493 -1 1 493 -1 1
248 264
MSE 41.33333333 44
MAPE
Month Demand Technique 1 Error Error Absolute (Error Absolute / Actual) x 100 Technique 2 Error Error Absolute (Error Absolute / Actual) x 100
1 492 488 4 4 0.008130081 495 -3 3 0.006147541
2 470 484 -14 14 0.029787234 482 -12 12 0.024793388
3 485 480 5 5 0.010309278 478 7 7 0.014583333
4 493 490 3 3 0.006085193 488 5 5 0.010204082
5 498 497 1 1 0.002008032 492 6 6 0.012072435
6 492 493 -1 1 0.00203252 493 -1 1 0.002028398
0.056319819 0.069829177
11/18/2023 54 3-54
Quiz – Chapter 3
3-55
Chapter 3
A forecast used for Master Production Scheduling and Control
is likely to cover a period of _____________. A few days to a
few weeks
Regression analysis where the relationship between variables
is a straight line is called _______ _______. Linear regression
In a time series analysis, time is the _________ variable.
Independent
An exponential smoothing forecast considers all past data
(T/F). True
In an exponential smoothing forecast, a higher level of alpha
(α) will place more emphasis on recent history (T/F). True
Mean error of a forecast provides information concerning the
forecast’s ________. Bias
3-56
Determining the Coefficient of
Correlation
Coefficient of Determination, r2, measures the percent of
change in y predicted by the change in x
Values range from 0 to 1
Easy to interpret
3-57