Scribd
Upload
31 views2 pages

A Note On Forecast Error

This document discusses different types of forecast errors including forecast error, mean absolute deviation, mean squared error, mean absolute percentage error, and tracking signal. It explains how to calculate and interpret each error type and highlights that the direction of tracking signal depends on the choice of error formula.

Uploaded by

Nikhil Kumar
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
31 views2 pages

A Note On Forecast Error

This document discusses different types of forecast errors including forecast error, mean absolute deviation, mean squared error, mean absolute percentage error, and tracking signal. It explains how to calculate and interpret each error type and highlights that the direction of tracking signal depends on the choice of error formula.

Uploaded by

Nikhil Kumar
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/ 2

A note on Forecast error

In class we have tried understanding various types of forecasting errors and the purpose
that one type of error calculation serves over the other. In this note, I am trying to present
a summary of our discussion and an interpretation of tracking signal. The note will walk
you through how these calculations have been reported in literature, in general and also
try to highlight the exceptions. Let’s understand them one by one:

Types of error:

1. Forecast error for a period t ,

et = (Actual Demand-Forecast Demand) = (At-Ft )

In some texts, you may find it represented as:


et=(Forecast Demand - Actual Demand) = (Ft - At)

Considering this, may cancel out the negative errors against the positive errors and the
net error may not be correct interpretation, and hence we go for next type of forecast.

2. Mean absolute deviation, MAD:

∑"
!#$ |#! |
MAD for n periods of demand data, MADn=
$

The absolute value takes care of the issue faced by error but fails to convey the direction
of the error.

3. Mean Squared Error, MSE:

∑"
!#$(#! )
%
Mean Squared Error for n periods, MSEn = $

However, if the error values are large, this magnifies the error part and becomes
cumbersome to handle.

4. Mean Absolute Percentage Error, MAPE:

|'! |
∑"
!#$ )!
Mean Absolute Percentage Error for n Periods, MAPEn = $

This allows us to view the error in absolute terms with respect to the actual demand
value.

1
5. Tracking Signal:

All the error calculations that we have seen so far fail to tell the direction of forecast error.
To overcome this, this metric is used:

'()*
Tracking Signal= +,-

Bias for n periods = ∑$./0 𝑒. and MAD values we know by now.

The direction of Tracking signal will purely depend on the ± 𝐵𝑖𝑎𝑠 value as MAD has a non-
negative value.

The catch here lies with the choice of error definition. Let’s go back to the error formula.

Case 1:
If , et = (Actual Demand-Forecast Demand) = (At-Ft )

And if At<Ft ,
we are over forecasting and the value of over forecasted value is represented by “-”

Also, if At>Ft ,
we are under forecasting and the value of under forecasted value is represented by “+”

Case 2:
However, if et = (Forecast Demand-Actual Demand) = (Ft-At )

And if At<Ft ,
we are over forecasting and the value of over forecasted value is represented by “+”

Also, if At>Ft ,
we are under forecasting and the value of under forecasted value is represented by “-”

And same is reflected on the final TS interpretation. So, the under or over forecast
interpretation lies with the choice of error formula.

As per most standard practice, we opt for case 1 and in that case “-” refers to over
forecast and “+“ to under forecast. So, don’t get confused if the signs of TS are changed
based on the choice of the error calculation. In class we used case 2, and hence the TS
interpretation was just the reverse.

For range of TS value within which the forecast is within control and other numerical
calculations, please refer to Page 395-397, Russel & Taylor , Eighth Edition

You might also like