Upload
piers-watson
View
226
Download
3
Tags:
Embed Size (px)
Citation preview
Problem Identification. Environmental analysis. Variable Identification. Forecasting. The Use of multiple models. Model categories. Model management. Knowledge-based modeling.
The perception of a difference between the current state of affairs and the desired state of affairs.
The problem statement contains (3) components:◦ Current state of affairs.◦ Desired state of affairs.◦ Central of objectives that distinguishes the two.
A common error in the formation of the problem is a premature focus on the choice set of solutions rather than the problem itself.
Problem Scope Once the problem is defined, the decision maker
must examine the scope of the problem, i.e. available resources, cognitive limitation, time constraints, etc
Decision-making under certainty: assumed that complete knowledge is available so that
decision maker knows exactly what the outcome of each course of action will be.
Decision-making under Uncertainty:Several outcomes are possible for each course of action. Decision maker does not know, or can't estimate the
probability of occurrence of the possible outcome. Decision-making under Risk: Several possible (random) outcomes for each action with
several probabilities. Risk analysis must calculated.
Influence Diagrams A method of modelling a decision
Sales volume
low
medium
high
A B
BA
A B
BA
A outcome is relevant to the probability of event B
Decision A is necessary to estimate probability of event B
Outcome of event A is known when making decision B
Decision A is made prior to decision B
Investment example
One goal: maximize the yield after one year
Yield depends on the status of the economy (the state of nature)◦ Solid growth◦ Stagnation◦ Inflation
1. If solid growth in the economy, bonds yield 12%; stocks 15%; time deposits 6.5%
2. If stagnation, bonds yield 6%; stocks 3%; time deposits 6.5%
3. If inflation, bonds yield 3%; stocks lose 2%; time deposits yield 6.5%
Decision variables (alternatives) Uncontrollable variables (states of economy) Result variables (projected yield)
States of Nature
Solid Stagnation Inflation
Alternatives Growth
Bonds 12% 6% 3%
Stocks 15% 3% -2%
CDs 6.5% 6.5% 6.5%
10
Optimistic approach
Pessimistic approach
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. Aronson, 6th editionCopyright 2001, Prentice Hall, Upper Saddle River, NJ
11
Use known probabilities (Table 5.3)
Risk analysis: compute expected values
Can be dangerous
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. Aronson, 6th editionCopyright 2001, Prentice Hall, Upper Saddle River, NJ
12
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. Aronson, 6th editionCopyright 2001, Prentice Hall, Upper Saddle River, NJ
Decision Tree A more detailed method of modelling a
decision
Enter contest
win contest win large
return of wager
Lose wager
Lose/gain nothing
loose contest
Do not enter contest
Basic Risky Decision Problems faced by the decision maker that
require a choice selection in the face of some uncertainty
Blasters soft drink example
uncertainty
Decision
Basic risky decision
Objective
Objective 1
Objective 2
Objective n
Total satisfactionDecision
uncertainty
Basic risky decision with multiple objectives
Certainty Decision Structure It involves situation in which the trade-off among the various
objectives and risk is not significant A variation of this structure is the multiple objectives and no-
risk decision It arises in situation that are so broad and complex
Objective 1
Objective 2
Objective n
Total satisfactionDecision
multiple objectives, no-risk decision
Sequential Decision Structure Conditions during the decision process may change over time,
and a choice made earlier may not be appropriate any more It represents a series of basic risky decisions in a successive
time period with arrows to indicate the relationship between each temporal set
Objective 1 Objective 2 Objective n
Total satisfactionDecision 1
multiple period sequential decision
Decision 2 Decision n
Uncertaintyt1
Uncertaintyt2
Uncertaintyt3
A model is a simplified representation of a real situation. modeling is the process of developing, analyzing and interpreting a model in order to help make better decisions.
Decision models can be classified in a number of ways, i.e. time, mathematical or logical focus.
A problem can be thought as a set of subsystems that are functionally decomposable at the desire of the decision maker
Abstract Model characteristics:1.It focuses on the mathematical precision with which various
outcomes can be predicted. 2.Since each subsystem of the problem context is modelled and
further decomposed, some detail of the information is lost to the decision maker
Abstract Model can be divided into four subsystems:
Deterministic Models: Stochastic Models Simulation Models Domain Specific Models
The construction of a simulation model for discrete events goes through the following steps:
State the objective of the model Define the scope and boundary of the system Define the state of the system Define all events that can effect the system state
and their individual impact on each state variable Define the unit of time used by the system Create statistical definition for each event in the
model Determine, a priori, the metrics desired from the
model Define the starting state of the model
A formal mathematical approach to a problem may not be the most appropriate strategy? Check out the disadvantages of Abstract Models ? Page 116.
Conceptual model can be thought as an analogies to the problem context
Experience from a past problem context can be used to assist in forecasting events and outcomes in a new context
Conceptual model often criticized as a subjective and individually biased toward the beliefs of the decision maker
Steve Hornik Drive up mail example (pages 118-119)◦ Strong acceptance◦ Poor acceptance◦ Moderate acceptance
A decision is not much of a decision unless some uncertainty is present
Uncertainty must be quantified in some manner in order for a decision to be made with any degree of success
Three requirement of probability:◦ All probability must be within the range of 0 to ◦ The probabilities of individual outcomes of an event must add
up to the probability of their union◦ The total probability of a complete set of outcomes must be
equal to 1
Types of probability◦ Long-run frequency◦ Subjective◦ Logical
Direct Forecasting
Odds Forecasting◦ It focuses on gambling prospective◦ The goal is to find a specific amount of money to
win or lose
Comparison Forecasting◦ Similar to the odds forecasting method◦ It presents the decision maker with a choice
between of one of lottery-like events