1. Management Decision Support and Intelligent Systems
Chapter 12
Management Decision Support and Intelligent Systems
Information Technology For Management 6th Edition
Turban, Leidner, McLean, Wetherbe
Lecture Slides by L. Beaubien, Providence College
John Wiley & Sons, Inc.
Chapter 12

2. DSS
DSS is a computer-based information system that contributes decision making process
Why do we need computers
Large number of alternatives (ex: FLP)
Uncertainty
Intensive calculations
Decision makers and data are usually located in different places
Decision making process can be
Highly structured (The procedures for obtaining the best solutions are known- Management Science)
Highly unstructured problems (managerial judgment and intuition; ex new product development)
Semi-structured Problems (DSS is most suitable)
Chapter 12

3. A model is a simplified representation of reality.
With modeling, one can reach an optimum solution under certain assumptions and perform virtual experiments
The cost of virtual experimentation is much lower
Years of operations can be simulated
Manipulating the model is much easier
What ifs
Evaluating large number of alternatives
Chapter 12

4. Optimization models LP (Lindo, QSB, AMPL) MIP (Lindo, QSB, AMPL)
Non-linear optimization models (Ex: Forecasting Models, SS calculation)
Analytical optimizatin models
Heuristic search algorithms
WinQSB vs AMPL
Chapter 12

5. Sample Ampl Code set iplnt := 1 2 3 4;
set kprod := ;
set tperiod := ;
set machine := ;
Subject to capacity {i in iplant, m in mach, t in tperiod}:
sum{k in kprod} production[i,k,m,t]/machrate[i,m,k]<= 28;
This code represents a set of 168 constraints (4 plants * 6 periods * 7 machines)
Chapter 12

6. Using Binary Variables
subject to c24 {t in 1..6, i in iplnt, j in jdest, k in kprod}:
fbbinary [i,j,k,t] + fcbinary [i,j,k,t] <= 1;
subject to c25 {t in 1..6, i in iplnt, j in jdest, k in kprod}:
cshipvar[i,j,k,t]<=300*fcbinary[i,j,k,t];
subject to c26 {t in 1..6, i in iplnt, j in jdest, k in kprod}:
bshipvar[i,j,k,t]>=301*fbbinary[i,j,k,t];
subject to c27 {t in 1..6, i in iplnt, j in jdest, k in kprod}:
bshipvar[i,j,k,t]<= *fbbinary[i,j,k,t];
subject to c28 {i in iplnt, j in jdest, k in kprod, t in 1..6}:
zshipment[i,j,k,t] = bshipvar[i,j,k,t] + cshipvar[i,j,k,t];
Chapter 12

7. When the model is solved, the results depend on these assumptions.
WHAT-IF ANALYSİS
A model builder makes predictions and assumptions regarding the input data, many of which are based on the assessment of uncertain futures.
When the model is solved, the results depend on these assumptions.
What-if analysis attempts to check the impact of a change in the assumptions (input data) on the proposed solution.
In a properly designed DSS, managers themselves can easily ask the computer these types of questions as many times as needed for sensitivity analysis.
Shadow price (WinQSB)
GOAL-SEEKING ANALYSIS
Goal-seeking analysis attempts to find the value of the inputs necessary to achieve a desired level of output.
It represents a “backward” solution approach.
For example, let us say that a DSS solution yielded a profit of $2 million.
Management wants to know what sales volume would be necessary to generate a profit of $2.2 million.
This is a goal-seeking problem.
Excel
Chapter 12

8. Stochastic optimization models
Ex: Forecast model fitting using solve in Excel
Chapter 12

9. Simulation
In DSSs, simulation generally refers to a technique for conducting experiments (such as “what-if”) with a computer on a model of a management system.
Because a DSS deals with semistructured or unstructured situations, it involves complex reality, which may not be easily analyzed by optimization or other standard models but often can be handled by simulation. (Less assumptions)
Simulation is increasingly part of the decision support tools used in the airline industry for short-term decision making
Chapter 12

10. Distribution assumption (statfit)
it can describe or predict the characteristics of a given system under different circumstances. Once the characteristics’ values are computed, the best among several alternatives can be selected. (ex. FLP)
The simulation process often consists of the repetition of an experiment many, many times to obtain an estimate of the overall effect of certain actions (Precision test).
Automod vs Promodel
Distribution assumption (statfit)
Chapter 12

11. Sample Automod code begin P_proc arriving if A_location = 1 then begin
move into Q_oakmachine /* processing queue */
use R_oakmachine for n 15, 240 min
inc V_production_cost(A_location) by 15
inc V_total_inv(A_location, A_loadindex) by 1
end
Chapter 12

12. Advantages of Simulation
Allows for inclusion of the real-life complexities of problems. Only a few simplifications are necessary. For example, simulation may utilize the real-life probability distributions rather than approximate theoretical distributions.
Is descriptive. This allows the manager to ask what-if type questions. Thus, managers who employ a trial-and-error approach to problem solving can do it faster and cheaper, with less risk, using a simulated problem instead of a real one.
Can handle an extremely wide variation in problem types, such as inventory and staffing, as well as higher managerial-level tasks like long-range planning. Further, the manager can experiment with different variables to determine which are important, and with different alternatives to determine which is best.
Chapter 12

13. Decision Complexity
Chapter 12

14. Characteristics and Capabilities of DSSs
Sensitivity analysis is the study of the impact that changes in one (or more) parts of a model have on other parts.
What-if analysis is the study of the impact of a change in the assumptions (input data) on the proposed solution.
Goal-seeking analysis is the study that attempts to find the value of the inputs necessary to achieve a desired level of output.
Chapter 12

15. DSS Process
Chapter 12

16. Intelligent Systems
Expert systems (ESs) are attempts to mimic human experts. It is decision-making software that can reach a level of performance comparable to a human expert in some specialized and usually narrow problem area. The idea is simple: expertise is transferred from an expert or other source of expertise to the computer.
The transfer of expertise from an expert to a computer and then to the user involves four activities:
Knowledge acquisition (from experts or other sources). It involves studying the thought processes of humans
Knowledge representation (organized as rules or frames in the computer)
Knowledge inferencing is performed in a component called the inference engine of the ES and results in the recommendation.
Knowledge transfer to the user (the expert’s knowledge has been transferred to users). It advises the nonexperts, and explaines, if necessary, the logic behind the advise.
Chapter 12

17. Artificial Neural Networks (ANN)
Pattern recognition and learning are the key characteristics of ANN
ANN can analyze large quantities of data to establish patterns where the logic rules are not known
Neural networks are particularly good at identifying subtle, hidden, and newly emerging patterns within complex data as well as interpreting incomplete inputs.
Example 1: Loan application. By reviving many historical cases of applicants’ questionnaires, ANN can create patterns or profiles of applicants that should be approved or denied
Example 2: Identifying patterns in stock market data and assisting in stock and bond trading strategies;
Chapter 12

18. Artificial Neural Network
Chapter 12

19. It is through repeated adjustments of weights that the network learns.
Inputs. Each input corresponds to a single attribute. For example, if the problem is
to decide on approval or disapproval of a loan, some attributes could be the applicant’s
income level, age, and home ownership. Several types of data, such as text,
pictures, and voice, can be used as inputs. Preprocessing may be needed to convert
the raw data to meaningful inputs from symbolic data or to scale the data. The
inputs are multiplied by weights—when they enter the processing elements (PEs).
Weights. Key elements in an ANN are the weights.Weights express the relative
strength (or mathematical value) of the input data or the many connections that
transfer data from layer to layer. In other words, in the case of a loan application,
weights express the relative importance of each input applicant’s attribute, based
on past experiences.Weights are crucial in that they store learned patterns of information.
It is through repeated adjustments of weights that the network learns.
Summation function. The summation function (represented by the symbol) calculates
the weighted sum of all the input elements entering each processing element.
A summation function multiplies each input value by its weight and totals
the values for a weighted sum.
Transformation function. A transformation function (represented by) integrates
the information produced by all PEs, and transforms it to meaningful outputs.
Outputs. The outputs of the network contain the solution to a problem. For
example, in the case of a loan application it can be “yes” or “no.”The ANN assigns
numeric values, like 1 for “yes” and 0 for “no.” The purpose of the network is to
compute the values of the output.
Chapter 12

20. Some Applications of Neural Networks
Data mining- Finding data in large and complex databases
Credit card fraud detection Analyzing purchasing patterns for fast detection of fraud
Tax fraud Identifying, enhancing, and finding irregularities
Evaluation of personnel and Matching personnel data to job requirements and performance criteria job candidates
Loan application evaluation- Judging worthiness of loan applications based on patterns in previous application information (customer credit scoring)
New product analysis- Sales forecasting and targeted marketing
Insurance fraud detection Finding fraud patterns (see hnc.com)
Chapter 12

21. Automated Decission Suppport (ADS)
ADS systems are rule-based systems that are particularly useful for repetitive managerial problems.
ADS is achieved by capturing a business user expertise in a set of business rules.
These rules determine what actions needs to be taken in particular situations
Examples
Loan approval
Dell
Yield optimization
Traffic lights
Dynamic Forecasting
Chapter 12

22. Why Managers Need IT Support
A key to good decision making is to explore and compare many relevant alternatives. The more alternatives that exist, the more computer-assisted search and comparisons are needed.
Typically, decisions must be made under time pressure. Frequently it is not possible to manually process the needed information fast enough to be effective.
It is usually necessary to conduct a sophisticated analysis in order to make a good decision. Such analysis requires the use of modeling.
Chapter 12

23. Managerial Issues
Cost justification, intangible benefits. While some of the benefits of management support systems are tangible, it is difficult to put a dollar value on the intangible benefits of many such systems.
Documenting personal DSS. Many employees develop their own DSSs to increase their productivity and the quality of their work. It is advisable to have an inventory of these DSSs and make certain that appropriate documentation and security measures exist.
Ready-made commercial DSSs. With the increased use of Web-based systems and ASPs, it is possible to find more DSS applications sold off the shelf, frequently online. The benefits of a purchased or leased DSS application sometimes make it advisable to change business processes to fit a commercially available DSS.
Embedded technologies. Intelligent systems are expected to be embedded in at least 20 percent of all IT applications in about 10 years. It is critical for any prudent management to closely examine the technologies and their business applicability.
Chapter 12