MECHANICAL ENGINEERING

DEPARTMENT Industrial Organization – MPD412

Lecture 3
Demand
Forecasting
Cont.

Prepared by : Dr. Sayed Ali Zayan 1st Term 2023/2024

Forecast Accuracy:
A forecast is never completely accurate; forecasts will always deviate from
the actual demand. This difference between the forecast and the actual is
the forecast error. Specifically,

where

Managers are usually more interested in measuring forecast error over a

relatively long period of time, however. The following are some of the more
commonly used methods:

Mean Absolute Deviation (MAD):

The smaller the value of MAD, the more accurate the forecast, although
viewed alone, MAD is difficult to assess. One benefit of MAD is to compare
the accuracy of several different forecasting techniques.
Forecast Accuracy:
Mean Absolute Percent Deviation (MAPD):
A lower percent deviation implies a more
accurate forecast.

Mean absolute percent error:

Cumulative sum of forecast errors:

Large positive value indicates that the forecast is probably consistently

lower than the actual demand, or is biased low. A large negative value
implies that the forecast is consistently higher than actual demand, or is
biased high.
The average the cumulative error over the number of time periods:
Example:
The manager of a large manufacturer of industrial pumps must
choose between two alternative forecasting techniques. Both
techniques have been used to prepare forecasts for a six
month period. Using MAD as a criterion, which technique has
the better performance record?
Solution:
Check that each forecast has an average error of approximately zero.

Technique 1 is superior in this comparison because its MAD is smaller,

Monitoring The Forecast
Many forecasts are made at regular intervals (e.g., weekly,
monthly, and quarterly). Because forecast errors are the rule
rather than the exception, there will be a succession of forecast
errors. Tracking the forecast errors and analyzing them can
provide useful insight on whether forecasts are performing
satisfactorily.
There are a variety of possible sources of forecast errors,
including the following:
 The model may be inadequate .
 Irregular variations may occur due to severe weather or other
natural phenomena, …etc.
 The forecasting technique may be used incorrectly, or the
results misinterpreted.
 Random variations.
i) Control Chart Method
A very useful tool for detecting non-randomness in errors is
a control chart. Errors are plotted on a control chart in the
order that they occur
i) Control Chart Method
In order for the forecast errors to be judged “in control” (i.e., random),
two things are necessary.
 One is that all errors are within the control limits.
 The other is that no patterns (e.g., trends, cycles, non-centered data)
are present.
 Both can be accomplished by inspection.
i) Control Chart Method
 To construct a control chart, first compute the MSE. The
square root of MSE is used in practice as an estimate of the
standard deviation of the distribution of errors. That is,
 This formula without the square root is known as the mean
squared error (MSE), and it is sometimes used as a measure
of forecast error. Then
 Control charts are based on the assumption that when
errors are random, they will be distributed according to a
normal distribution around a mean of zero. Recall that for a
normal distribution, approximately 95.5 percent of the
values (errors in this case) can be expected to fall within
limits of 0 ± 2s (i.e., 0±2 standard deviations), and
approximately 99.7 percent of the values can be expected
to fall within ±3s of zero. With that in mind, the following
formulas can be used to obtain the upper control limit
(UCL) and the lower control limit (LCL):
Where: z = Number of
 Combining these two formulas, we obtain the standard deviations from
following expression for the control limits: the mean
ii) Tracking Signal (TS) method
 Another method is the tracking signal
(TS). It relates the cumulative forecast
error(CFE) to the average absolute error
(i.e., MAD), used to monitor a forecast.
The tracking signal is computed period
by period using the following formula:

 The intent is to detect any bias (Bias

Persistent tendency for forecasts to be
greater or less than the actual values of
a time series.), in errors over time (i.e., a
tendency for a sequence of errors to be
positive or negative).
 TS is a measure that indicates whether a
method of forecasting has any built-in
biases over a period of time.
ii) Tracking Signal (TS) method
 TS measures the number of MADs represented by the
cumulative sum of forecast errors.
 Values can be positive or negative.
 Positive tracking signals TS indicate that demand is
greater than forecast.
 Negative signals mean that demand is less than
forecast.
 A value of zero would be ideal;
 limits of ±4 or ±5 are often used for a range of
acceptable values of the tracking signal.
 If a value outside the acceptable range occurs, that
would be taken as a signal that there is bias in the
forecast, and that corrective action is needed
ii) Tracking Signal (TS) method
The tracking signal is recomputed each period, with
updated, “running” values of cumulative error and MAD.
The movement of the tracking signal is compared to
control limits; as long as the tracking signal is within these
limits, the forecast is in control.

After an initial value of MAD has been determined, MAD

can be updated and smoothed (SMAD) using exponential
smoothing:

Typically, forecast errors are normally distributed, which

results in the following relationship between MAD and the
standard deviation of the distribution of error, S:
Example:
Monthly attendance at financial
planning seminars for the past 24
months, and forecasts and errors for
those months, are shown in the following
table. Determine if the forecast is
working using these approaches:
1. A tracking signal, beginning with
month 10, updating MAD with
exponential smoothing. Use limits of ±4
and α = .2.
2. A control chart with 2s limits. Use data
from the first eight months to develop
the control chart, and then evaluate the
remaining data with the control chart.
Solution
1. The sum of absolute errors
through the 10th month is 58.
Hence, the initial MAD is
58/10 = 5.8.
The subsequent MADs are
updated using the formula
MADnew = MADold + α (|e| - MADold ).
The results are shown in the
following table.

The tracking signal for any month is

Solution

Because the tracking signal is within ±4 every month, there is no

evidence of a problem.
Solution
2. a. Make sure that the average error is approximately zero, because a
large average would suggest a biased forecast.

b. Compute the standard deviation:

c. Determine 2s control limits:

Solution
d.
(1) Check that all errors are within the limits. (They are.)
(2) Plot the data (see the following graph), and check for nonrandom
patterns. Note the strings of positive and negative errors. This suggests non-
randomness (and that an improved forecast is possible). The tracking signal
did not reveal this.

A plot helps you to visualize the process and enables you to check for
possible patterns (i.e., non-randomness) within the limits that suggest an
improved forecast is possible.
 Causal Forecasting models
(linear-regression analysis)
Linear regression is a mathematical technique that relates one
variable, called an independent variable, to another, the dependent
variable, in the form of an equation for a straight line.
A linear equation has the following general form:

Where
Y = computed value of the variable to be predicted
(called the dependent variable)
X = the independent variable (which is time in this case)
a = y-axis intercept
B = the slop
linear-regression analysis

Or
linear-regression analysis
Correlation coefficient for regression lines (r):
 The correlation coefficient measures the direction and strength of
the linear relationship between the independent variable and the
dependent variable.
 The correlation coefficient can be calculated using the following
equation:

 The coefficient of correlation (r) explains the relative importance of

the relationship between y and x.
 The sign of r shows the direction of the relationship, and the absolute
value of r shows the strength of the relationship,
Linear-regression analysis
 r can take any value between -1 to +1.
 The sign of r is always the same as the sign of b.
 negative of r indicates that the values of y and x tend to move in
opposite directions, and positive move in the same direction.
linear-regression analysis
The coefficient of determination r2:
 It measures the amount of variation in the dependent variable that
is explained by the regression line.
 It can be calculated with the following equation:

 The value of r2 ranges from 0.0 to 1.0.

 Regression equations with a value of r2 close to 1.0 are desirable
because the variations in the dependent variable and the forecast
generated by the regression equation are closely synchronized.
linear-regression analysis
Standard error of the estimate Syx
To measure the accuracy of the
regression estimates, we need to
compute the standard error of the
estimate. This is called the Standard
deviation of the regression.

 Syx measures how closely the

data on the dependent variable
cluster around the regression line.
or is a measure of how historical
data points have been dispersed
about the trend line.
Example:
City government has collected the following data on annual sales tax
collections and new car registrations:

Determine the following:

a) The least-squares regression equation
b) Using the results of part (a), find the estimated sales tax collections if
new car registrations total 22,000.
c) The coefficients of correlation.
Solution:
a) The least-squares regression equation
Solution:

The least-squares regression equation is:

b) Using the results of part (a), find the estimated sales tax collections if
new car registrations total 22,000.

c) The coefficients of correlation.