Professional Documents
Culture Documents
Demand Management Forecasting Causal Factors
Demand Management Forecasting Causal Factors
Demand
management
forecasting
Causal Factors
Amit Ben Zvi
Introduction .................................................................................................. 4
Conclusion ................................................................................................... 8
The Demand management (DM) analytical engine has various analytical settings to it, configured as part of a
“Forecasting Profile”.
The seeded forecasting profiles are fixed, and cannot be modified, but other, user-defined, profiles have various
configuration options for them, like configuring the forecast tree structure, selecting various forecasting
methods, forecasting parameters and causal factors.
This paper will focus on the decomposition groups and causal factors. This paper will describe and explain how
to configure them.
Historical Demand
Explanatories that can explain the historical demand – Causal factors.
Causal factors provide information about historical events that are expected to reoccur in the future.
Causal factor is a time-varying quantity that explains and drives demand and causes demand to deviate from a
norm.
A causal factor differs from any other historical explanatory by having full information not just in historical time
periods, but also in future dates.
Explanatories like Exchange rate or weather have full disclosure in history, but not any certainty for the future
values, as such, they are not good candidates to be a causal factor.
Causal factors are measures; new causal factors should be created as a measure, added to the relevant
catalog and the Forecasting profile.
The values for the measure (causal factor) should be defined for the complete time span of the plan (history and
future).
- January – December – defined on Time granularity only with a SUM time aggregation.
- Price – defined on Item and Time granularity only with a WAVG Aggregation and time aggregation.
Causal groups (Short, Long, Multiplicative, Not-Seasonal) define which forecasting methods will use this causal
factor.
As written in the analytical consideration section, models that can handle a larger set of casuals will use the
Long Causal group (models like Monte Carlo Regression).
Fill Causals parameter is relevant only for causals defined on Item – location – date granularity – defines if the
Analytical engine will automatically fill in data when there are missing dates – linear interpolation of the nearest
neighbors’ values, or use the value of Zero for missing dates.
Causal factor data should be as complete as possible. Otherwise, the following may occur:
o If a causal factor does not have values for future dates, it may not have desired effect on
forecasts
o Likewise, if the historical data is not long enough to learn the influence of all causal factors,
the forecast may generate unexpected results. The causal should be aligned with the
historical data it needs to explain.
If a causal factor has all zeros in the historical time period of the combination, the causal is ignored.
The forecasting methods need to have some ratio between the number of historical demand points and
the number of causal factors.
The analytical reasoning is to avoid over fit that can compromise the validity and accuracy of the
forecast.
Some forecasting methods handle a larger set of causal factors better than others; please refer to the
section that discusses the definition of causal factors for more information.
An example for Co-linearity in the seeded causals is the Constant causal factor that is the sum of the
Months causal factors (January – December) or the sum of the days of the week causals (Monday –
Sunday).
To avoid this seeded Co-linearity, the March and Tuesday causals are turned off.
It does not mean they are not being evaluated, but the weight of March and Tuesday (as in the seeded
configuration) is attributed to the Constant causal factors as not being any other month/ day.
The forecasting methods can tolerate some level of Co-linearity. In addition some Forecasting methods
can deal better with Co-linearity like Ridge and Monte Carlo regression (– the forecasting methods that
use the Long causal group), while most Forecasting methods, like Regression, are more sensitive to
Co-linearity.
DECOMPOSITION GROUPS
Each causal factor can belong to one Decomposition group, when adding a new causal it can be assigned to
one of the pre-configured Decomposition groups or assigned to a user-defined Decomposition group.
Decomposition group allows the user to split the total forecast to logical components, learning the effect of each
one of them.
* Enabling Decomposition affects the Engine run time performance and is not advisable for full batch engine
runs.
Causal factors are available in the Decomposition Tab in the Forecasting Profile screen.
Level Decomposition Group Constant Mandatory causal factors, that provides the base
demand.
Holidays Decomposition Group Christmas Holiday 1 when it is Christmas Holiday, 0 when it is not
Christmas Holiday
Price Decomposition Group Final Price Final Price measure, this causal can cause
performance issues in run time, so advisable to
disable it if not needed.
CONCLUSION
This document provided information about Causal factors and Decomposition groups used by the DM analytical
engine; it explains what Causal factors are, and how to create and use them.
Worldwide Headquarters
500 Oracle Parkway, Redwood Shores, CA 94065 USA
Worldwide Inquiries
TELE + 1.650.506.7000 + 1.800.ORACLE1
FAX + 1.650.506.7200
oracle.com
CONNECT W ITH US
Call +1.800.ORACLE1 or visit oracle.com. Outside North America, find your local office at oracle.com/contact.
Copyright © 2019, Oracle and/or its affiliates. All rights reserved. This document is provided for information purposes only, and the contents hereof are
subject to change without notice. This document is not warranted to be error-free, nor subject to any other warranties or conditions, whether expressed
orally or implied in law, including implied warranties and conditions of merchantability or fitness for a particular purpose. W e specifically disclaim any
liability with respect to this document, and no contractual obligations are formed either directly or indirectly by this document. This document may not be
reproduced or transmitted in any form or by any means, electronic or mechanical, for any purpose, without our prior written permission. This device has
not been authorized as required by the rules of the Federal Communications Commission. This device is not, and may not be, offered for sale or lease,
or sold or leased, until authorization is obtained. (THIS FCC DISLAIMER MAY NOT BE REQUIRED. SEE DISCLAIMER SECTION ON PAGE 2 FOR
INSTRUCTIONS.)
Oracle and Java are registered trademarks of Oracle and/or its affiliates. Other names may be trademarks of their respective owners.
Intel and Intel Xeon are trademarks or registered trademarks of Intel Corporation. All SPARC trademarks are used under license and are trademarks or
registered trademarks of SPARC International, Inc. AMD, Opteron, the AMD logo, and the AMD Opteron logo are trademarks or registered trademarks
of Advanced Micro Devices. UNIX is a registered trademark of The Open Group. 0419
White Paper Title
January 2017
Author: Dina Jacobs
Contributing Authors: [OPTIONAL]