Supply Chain and Decision Modeling Final
Two approaches to clustering discussed in the text are
k-means clustering and hierarchical clustering
The PsiTarget(.) function in Analytic Solver Platform
returns the cumulative probability of a specified distribution cell being less or equal to the specified target value
Consider the constraint:X1+(d-)-d+= 5. Suppose that X1 = 3 in the optimal solution. The values of deviational variables and are:
see image for answer
Before effectively applying the k nearest neighbor classification technique, the variables need to be
standardized
Neural networks technique attempts to learn
what relationship exists between a set of input and output variables
In order to indicate the output cell (or cells) that we want Analytic Solver Platform (ASP) to track during the simulation we can use the:
"Add Output" button on the Analytic Solver Platform (ASP) menu.
Sensitivity analysis is useful when:
You may be interested in examining how sensitive the simulation output results are to various uncertain input cells in the model.
A MINIMAX objective is sometimes helpful in goal programming (GP) when:
You want to minimize the maximum deviation from any goal.
Refer to Exhibit 14.3. What is the expected monetary value of Investment A?
a. 20 (add total)
The decision tree is:
a. A graphical presentation of the information available in the payoff table. b. Intuitive to use. c. Easy to use for multi-stage decisions. (All of the above are correct)
Simulation is:
a. A technique that measures various characteristics of the model bottom-line performance measure. b. A technique that describes various characteristics of the bottom-line performance measure. c. A technique that is useful when one or more values for the independent variables are uncertain. (All of the above)
Data mining tasks fall into the following potential categories:
a. Classification. b. Prediction. c. Association/Segmentation. ( All of the above are categories of data mining tasks)
The RHS value of a goal constraint is referred to as the
Target Value
If we don't know what value a particular cell in a spreadsheet will assume and enter a number that we think is the most likely value for the uncertain cell, we can calculate the most likely value of the bottom-line performance measure. This is called:
The base-case scenario.
When faced with a classification problem, careful consideration should be given to:
The composition of the training sample.
The term 'iterative solution procedure' means that:
The decision maker investigates a variety of solutions to find one that is most satisfactory.
Suppose that the regrets for an alternative with three states of nature are: 20, 10, and 0. The probabilities of these states of nature are 0.2, 0.3, and 0.5, respectively. The expected regret for the alternative is equal to
The expected regret is: 20(0.2) + 10(0.3) + 0(0.5) = 7
Tornado charts in Analytic Solver Platform (ASP) help identify:
The inputs that have the greatest impact on the EMV
The challenge with data availability today is getting:
The right data in the right amount for the problem at hand.
Under probabilistic decision rules:
The states of nature in a decision problem can be assigned probabilities of occurrence.
Suppose that the objective function for a GP problem is: MIN . The term ti represents: (see image)
The target value for goal i.
A random variable is a:
Variable whose value cannot be predicted or set with certainty.
Affinity analysis is:
A data mining technique aimed at discovering what goes with what.
Cluster analysis is:
A data mining technique used to identify meaningful groupings of records within a data set.
In what-if analysis:
A manager changes the values of the uncertain input variables to see what happens to the bottom-line performance measure
A table that summarizes the final outcome (or payoff) for each decision alternative under each possible state of nature is referred to as:
A payoff matrix.
Suppose that all goal constraints in a goal programming problem are hard and the objective is: MIN (see pic)
All goals must be met exactly
The decision maker has expressed concern with Goal 1, budget, achievement. He indicated that future candidate solutions should stay under budget. How can you modify your goal programming model to accommodate this change?
All: a. Make budget a hard constraint in the model. b. Give d1+ an extremely large weight in the objective function. c. Remove d1+ from the goal constraint.
The expected monetary value (EMV) and expected opportunity loss (EOL) decision rules:
Always result in the selection of the same decision alternative
A course of action intended to solve a problem is called:
An alternative
Suppose that in goal programming (GP) we assign arbitrarily large weights to deviations from these goals to ensure that undesirable deviations from them never occur. This approach is called:
. Preemptive GP.
When such a decision is made, some chance exists that the decision will not produce the intended results. This chance, or uncertainty, represents:
. Risk
Given the following confusion matrix what is the correct classification rate? (actual (of group 1)+group (of group 2)/total (of 1&2)
19/25 = 76%
Suppose the highest payoff decision for a given state of nature is 100. You made a decision with a payoff of 80. The regret (or opportunity loss) for your decision is:
20.
Suppose that the first goal in a GP problem is to make 3 X1 + 4 X2 approximately equal to 36. Using the deviational variables d1− and d1+, what constraint can be used to express this goal?
3 X1 + 4 X2 + d1− − d1+ = 36
Refer to Exhibit 7.3. What formula goes in cell E11? (weighted % deviation)
=D11*(C11−B11)/C11
Using the information in Exhibit 12.3, what formula should go in cell H8 to determine if an order should be placed?
=IF(G8<$D$3,1,0)
Using the information in Exhibit 12.3, what formula should go in cell J8 to determine the arrival date for an order?
=IF(I8=0,0,A8+1+I8)
Refer to Exhibit 14.3. What formula should go in cell F5 and copied to F6:F8 of the spreadsheet if the expected regret decision rule is to be used?
=MAX(B$5:B$8)-B5
Using the information in Exhibit 12.3, what Analytic Solver Platform function should be used in cell I8 to determine the lead time for an order?
=PsiDiscrete($B$6:$B$8, $C$6:$C$8)
Using the information in Exhibit 12.3, what Analytic Solver Platform function should be used for generating a random shipping time based on the Data spreadsheet distribution for shipping time?
=PsiDiscrete($B$6:$B$8, $C$8:$C$8)
Using the information in Exhibit 12.3, what Analytic Solver Platform function should be used in cell D8 and copied to cells D9:D21 of the MODEL sheet to compute daily demand?
=PsiDiscrete($E$6:$E$13, $F$6:$F$13)
Which Analytic Solver Platform function will generate random integer numbers between 2 and 8?
=PsiIntUniform(2, 8)
Logistic regression is:
A classification technique that estimates the probability of an observation belonging to a particular group.
Refer to Exhibit 14.3. What decision should be made according to the expected monetary value decision rule?
C ( Highest of prob. totals)
When a credit manager of a mortgage company identifies the loans as those resulting in default and those that are current, he/she uses:
Classification.
Overfitting refers to a situation when the tree algorithm:
Classifies new observations less accurately than trees that do not overfit the training data.
The first step to create the training and validation data set using XLMiner Platform is:
Click the Partition icon in the Data Mining section.
In Discriminant Analysis, the F1 score:
Combines the precision and recall measures to provide an overall measure of a classifier's accuracy.
The first step in using the minimax regret decision rule is:
Converting the payoff matrix into a regret matrix.
Analytic Solver Platform (ASP) provides several "Psi" functions that can be used to:
Create the RNGs required for simulating a model.
The ______ in a decision problem represent various factors that are important to the decision maker.
Criteria.
Refer to Exhibit 7.3. Which value should the investor change, and in what direction, if he wants to reduce the risk of the portfolio?
D12, increase (Average risk)
Refer to Exhibit 14.3. What decision should be made according to the expected regret decision rule?
Decision C should be made according to the expected regret decision rule.
Classification techniques differ from most other predictive statistical methods, such as regression analysis, because the dependent variable is:
Discrete.
Goal programming solution feedback indicates that the d4+ level of 50 should not be exceeded in future solution iterations. How should you modify your goal constraint40 X1 + 20 X2 + d4− + d4+ = 300to accommodate this requirement?
Do not modify the constraint, add a constraint d4+ ≤ 50
The term "data mining":
Encompasses a variety of analytic techniques that can be used to help managers analyze, understand, and extract value from large sets of data.
Goal programming (GP) provides a way of analyzing potential solutions to a decision problem that involves soft constraints. Soft constraints can be stated as:
Goals with target values and deviational variables, which measure the amount by which a given solution deviates from a particular goal.
Cannot be violated
Hard Constraint
The goal of decision analysis is to:
Help individuals make good decisions.
In Discriminant Analysis (DA), precision is a measure of:
How accurate the classifier is when it predicts a "success".
The k-nearest neighbor (k-NN) technique:
Identifies the k observations in the training data that are most similar (or nearest) to a new observation we want to classify.
Goal programming:
Involves solving problems containing a collection of goals that we would like to achieve.
The decision rule that selects the alternative associated with the largest payoff is:
Maximax
The decision rule that pessimistically assumes that nature will always be "against us" regardless of the decision we make is:
Maximin.
Suppose that you have an option to hire a consultant who has the ability to predict the future with 100 percent accuracy. Using the consultant's reliable recommendations, you found that the expected value with perfect information is equal to $200. Without the consultant's insights you determined the EMV to be equal to $175. Would you pay the consultant $30 for her service? Why or why not?
No, because EVPI is $25, which is less than the consultant's fee of $30.
The decision rules that assume that probabilities of occurrence are not known or cannot be assigned to the states of nature in a decision problem are referred to as:
Nonprobabilistic rules.
The first step in performing a simulation in a spreadsheet is:
Placing a random number generator (RNG) formula in each cell that represents a random, or uncertain, independent variable.
The manager hopes that using the expected, or most likely, values for all the uncertain variables will:
Provide the most likely value for the bottom-line performance measure (Y)
The purpose of discriminant analysis (DA) is to:
Provide theoretically optimal or good classification results.
The key concept in goal programming (GP) is:
Reconciling trade-offs between conflicting goals.
Recalculating the spreadsheet several hundred or several thousand times and recording the resulting values generated for the output cell(s), or bottom-line performance measure(s) is called:
Replication.
Deviational variables:
Represent the amount by which each goal deviates from its target value.
We can apply a process known as ______ to a decision tree to determine the decision with the largest EMV.
Rolling back
Your company decided to build a manufacturing plant in Georgia in anticipation of increased demand for the product it produces. The level of demand in this problem is a(n):
State of nature
Future events that are not under the decision maker's control are known as:
States of nature.
Numeric constants that can be assigned values to weight the various deviational variables in the problem
Suppose that the objective function for a GP problem is: MIN: see pic, The terms W- & W+ and represent:
A classification tree:
a. Is a graphical representation of a set of rules for classifying observations into two or more groups. b. Use a hierarchical sorting process consisting of splitting nodes and terminal nodes to group records from a data set into increasingly homogeneous groups. c. Are popular because the resulting classification rules are very apparent and easy to interpret. (All of the above are characteristics of classification trees)
A strategy table:
a. Is a sensitivity analysis technique. b. Allows a decision maker to analyze optimal decision strategy changes. c. Allows two simultaneous changes in probability estimates. (All of the above answers)
When faced with uncertainty, people do the following:
a. React with paralysis. b. Do exhaustive research. c. Avoid making a decision. (All of the above)
Neural networks:
a. Simulate human learning. b. Are computer programs modeled after computing architecture of the human brain. c. Are a pattern recognition technique that attempts to learn what relationship exists between a set of input and output variables. (All options are correct)
The benefit(s) of simulation include(s):
a. The results of simulation do give us greater insight into the problem. b. It gives a decision maker some idea of the best- and worst-case total outcomes for the problem. c. It provides an idea of the distribution and variability of the possible outcomes. (All of the above)
A company is planning a plant expansion. They can build a large or small plant. The payoffs for the plant depend on the level of consumer demand for the company's products. The company believes that there is an 69% chance that demand for their products will be high and a 31% chance that it will be low. The company can pay a market research firm to survey consumer attitudes towards the company's products. There is a 63% chance that the customers will like the products and a 37% chance that they won't. The payoff matrix and costs of the two plants are listed below. The company believes that if the survey is favorable there is a 92% chance that demand will be high for the products. If the survey is unfavorable there is only a 30% chance that the demand will be high. The following decision tree has been built for this problem. The company has computed that the expected monetary value of the best decision without sample information is 154.35 million. What is the EVSI for this problem (in $ million)?
b. 0.07
A payoff matrix depicts ____ versus ____ with payoffs for each intersection cell.
decision alternatives; states of nature
In using neural networks, an analyst must decide __________ and ___________
how many hidden layers to use and how many nodes to use in each of the hidden layers
