Application of the Analytic Hierarchy Process AHP for Selection of Forecasting Software

1
Application of the Analytic Hierarchy Process
(AHP) for Selection of Forecasting Software

• Altug Pekin, Gamze Ozkan, Onur Eski, Umut
  Karaarslan,
• Gurdal Ertek, Kemal Kilic

2
Introduction

• Forecasting
• A fundamental activity carried out by almost
  every company

3
Introduction

• There is
• - A dramatic increase the number of products
  produced and sold in recent decades
• - An inevitable need for forecasting software
• Crucial decision
• Choosing the appropriate software

4
Introduction

• Our Study
• Selecting the best forecasting software with the
  help of AHP (Analytic Hierarchy Process)

5
Introduction

• Outline
• Introduction to AHP
• Filtering Software Products
• Deciding on the Criteria
• AHP using Expert Choice software
• Findings and Insights
• Sensitivity Analysis
• Related Literature
• Conclusions

6
The Analytic Hierarchy Process (AHP)

• AHP Model
• Problem is structured as a hierarchy
• Reflects the decision problems major components
  (decision criteria) and their inter-connections
  (comparisons with each other)
• In our problem, we have a single-level hierarchy
  of decision criteria

7
The Analytic Hierarchy Process (AHP)

• Comparisons
• A judgment or comparison is the numerical
  representation of a relationship between two
  elements that share a common parent (Saaty,
  1994)
• The judgments made in a scale ranging from 1 to 9
  

8
The Analytic Hierarchy Process (AHP)
9
The Analytic Hierarchy Process (AHP)

• Consistency
• One makes redundant comparisons to improve the
  validity of the answer.
• Redundancy gives rise to multiple comparisons of
  an element with other elements and hence to
  numerical inconsistencies. (Saaty, 1994)
• Inconsistency is tolerable if does not exceed 10.

10
The Analytic Hierarchy Process (AHP)

• AHP is distinguishable from its alternatives
• Even though constructing an AHP model requires
  eliciting of extensive data from a group of
  respondents, and is thus time consuming in this
  respect, it is fairly insensitive to judgmental
  errors. (Karlsson, 1998)

11
Deciding on the small subset of software

• 1st Step
• Compiling evaluations of forecasting software
  that were submitted as a part of a course project
• Among 100 projects, 48 of them (that received
  gt 40 out of 50) taken into consideration
• A list of 24 software products generated

12
Deciding on the small subset of software

• 2nd Step
• Another evaluation based on popularity by the
  groups and the availability of the trial versions
  on the Internet
• The obtained software list reduced to a smaller
  list of 13 software products

13
Deciding on the small subset of software

• 3rd Step
• 13 software products filtered through testing on
  a dataset according to
• Ease of use
• Steepness of the learning curve
• Suitability
• The list reduced to six products

14
Exhibits Trend and Strong Seasonality
15
Deciding on the Criteria

• 4th Step We identified seven criteria
• Six of these criteria derived from Tashman and
  Hoover (2001)
• Data preparation
• Method selection
• Method implementation
• Method evaluation
• Assessment of uncertainty
• Forecast presentation
• The seventh criterion Ease of use
• Selected as the most important feature by
  managers (selected by 86 of them) (Sanders and
  Manrodt, 2003)

16
Deciding on the Criteria

• Some criteria that were not included
• Price
• Country of Origin
• Import/Export capabilities

17
Application of the AHP using Expert Choice

• 5th Step Comparing the criteria with each other
• Comparing each software according to each
  criteria (7 matrices)
• Entering these matrices as input to Expert Choice

18
Our Findings and Insights

• The Priorities Computed by Expert Choice Software
  and the Values of Criteria Excluded from the AHP
  Model

19
Our Findings and Insights
20
Our Findings and Insights

• DecisionPro software, the best with respect to
  Method selection
• NCSS ranks as the top software product
• NCSS is shows superiority with respect to
  Forecast precision criterion, with the Aura
  software
• Forecasting Tools software
• The lowest-price alternative
• Highest score with respect to the Ease of use
• Worse with respect to other criteria
• Minitab software advantageous with respect to
  Uncertainty assessment
• Aura, in Russia

21
Sensitivity Analysis
22
Sensitivity Analysis
23
Related Literature

• Tashman and Hoover (2001)
• General insights and suggestions
• Ratings of the software products and the
  categories
• Omitted criteria such as ease of learning, and
  easy of use by decision makers who possess only a
  modest statistical background
• Our study takes as audience the decision maker
• With less technical knowledge
• Who has limited time to test various software

24
Related Literature

• Application of the AHP to selection of software
• Ossadnik and Lange (1999) Evaluate three AHP
  software products through an AHP-based study
• Lai et al. (2002) A case study that six software
  engineers participated, which involved selection
  of a multi-media authorizing system
• Post-study survey revealed that the AHP was more
  preferable than Delphi as a group-decision making
  method.
• Jung and Choi (1999) use AHP to derive weights of
  software modules based on access frequencies of
  the modules- that are then used in optimization
  models.

25
Conclusion

• Introduced the use of the AHP to the forecasting
  literature for the first time, to our knowledge.
• Insights with respect to which software products
  would be most appropriate for which types of
  companies
• Sensitivity analysis show that the weights given
  to decision criteria can change the priorities
  and rankings of the software products.

26
Questions?