Scribd
Upload
18 views18 pages

Forecasting

Uploaded by

ghz
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
18 views18 pages

Forecasting

Uploaded by

ghz
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 18

Forecasting

Forecasting: the art and science of


predicting future events
 Short term
 Medium term
 Long-term

Forecasting in planning future operations:


 Economic forecasts
 Technological forecasts
 Demand forecasts
Forecast methods:
 Qualitative strategies: decision maker’s
intuition, emotions, personal experiences…etc
 Quantitative techniques: time series models
(future = function of past), associative models
(future depends on factors or variables)
Time Series Models
Naïve approach

Ex: j: current period,


yj = sales at period j = 50 units
predict yj+1 = yj = 50
Moving average

moving average 
 (demand in previous n period )
n
The following is the report of monthly sales of ABC Company.

Month Actual sales 3-month moving average


Jan 10
Feb 12
Mar 13
Apr 16 ?
May 19 ?
Jun ?

Use 3-month average to predict the sales in Apr, May and June.
Exponential smoothing

Ft  (1   )Ft  1  At  1  Ft  1   ( At  1  Ft  1 )
Ft  new forecast,   smoothing constant (0  1),
Ft  1  previous forecast, At  1  actual demand in the previous period

Ex: In Jan, a car dealer predicted that Feb demand


would be 142 cars. The actual demand in Feb
is 153. Use a smoothing constant 0.2 to get the
forecast for March.
Ex: During the past 8 quarters, the Port of Baltimore has unloaded large
quantities of grain from ship (see the date below). The operations manager
suggested the use of exponential smoothing to predict tonnage unloaded.
The forecast of grain unloaded in the first quarter was 175 tons. Two values
of smoothing constant are used: 0.1 and 0.5. Perform the forecasting task
for the manager for the coming quarter. Also show the forecast against the
actual tonnage unloaded for the past 8 quarters.
Quarter Actual tonnage Forecast for  = Forecast for  =
unloaded 0.1 0.5
1 180
2 168
3 159
4 175
5 190
6 205
7 180
8 182
9 ? ?
Measuring Forecast Error:
Comparing the forecasted values using any
forecasting methods with the actual or observed
values.
Measures are often used in practice to calculate
the overall forecast error:
- Mean Absolute Deviation (MAD)
- Mean Squared Error (MSE)
- Mean Absolute Percentage Error (MAPE)
For the previous example (the Port of
Baltimore), determine which smoothing constant
(0.1 or 0.5) is better under the criteria of MAD,
MSE and MAPE, respectively.

forecast error actual demand - forecast value


n the number of periods the overallerror measured upon

MAD  | forecast error each period |


n
2
MSE 
( forecast error each period )
n
 100 | forecast error each period |
MAPE  actual %
n
Trend Projection approach:
ABC restaurant has kept track of the orders of the new
dinner special in the past seven weeks, see below:

Week Demand (no. of orders)


1 74
2 79
3 80
4 90
5 105
6 142
7 122
Use time projection to predict the number of orders for the new
dinner meal in week 8.
Question: In the example above, what’s the forecasted demand of week 8 by
using
• Naïve approach
• 5-week moving average?
Trend projection
Use “least square method” (introduced in the
statistics course, linear regression model) to find
the line, linear trend, over time
ˆ a  bx
y
where x is the time period and ŷ is the predicted
quantity.
The value of R2 (coefficient of determination) for
the trend line implies the usefulness of the line
indicating the variation of the output (Y)
explained by the input (X)
Seasonal variation in data
“Trend projection” predicts the trend of data
over a series of time. The adjustments may be
needed on it to fit seasonality.

Ex: The following table shows the demand of PC


each month in 2009, 2010 and 2011 (in
thousand). The demand for 2012 is expected
to be 1200 thousands. Use seasonal model to
predict the demand of each month in 2012.
Month 2009 2010 2011

Jan 80 85 105

Feb 70 85 85

Mar 80 93 82

Apr 90 95 115

May 113 125 131

Jun 110 115 120

Jul 100 102 113

Aug 88 102 110

Sep 85 90 95

Oct 77 78 85

Nov 75 82 83

Dec 82 78 80
past 36-month data
140

130

120

110

past 36-month data


100

90

80

70

60
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36
yearly data
1250

1200

1150

yearly data
1100

1050

1000

950
1 2 3
Multiplicative seasonality model
Step 1: Find the average historical demand each
season and the average demand over all
seasons
Step 2: Compute seasonal index for each season if
it is not given.
Step 3: Seasonal forecast = (seasonal
index)(trend forecast)
Complete the table

Month Step 1: average Step 2: overall Seasonal index Step 3: forecasts


monthly Seasonal demand of 2012
demand from
09~11
Jan

Feb

Mar

Apr

May

Jun

Jul

Aug

Sep

Oct

Nov
Associate Forecasting Method
Find the factor(s) which impact the output and use regression
analysis to forecast
Ex: A marketing research firm in Poughkeepsie has found the local
payroll is considerably associated with the sales of cars in the
town. They have collected the past 6 year data of annual sales and
local payroll as follows. Use associate forecasting to predict the
sales of this year if the local payroll of this year is expected to be 6
billion.

Sales (in million): 2.0 3.0 2.5 2.0 2.0 3.5

Local payroll (in billion): 1 3 4 2 1 7

You might also like