ISYE 6501 Final Quiz Actual Exam 2026/2027: Complete
Exam-Style Questions with Detailed Rationales | 100%
Verified | Pass Guaranteed – A+ Graded
TABLE OF CONTENTS
Section 1 | Classification Models | Q1 – Q10
Section 2 | Clustering and Dimensionality Reduction | Q11 – Q20
Section 3 | Time Series and Forecasting | Q21 – Q30
Section 4 | Validation and Model Selection | Q31 – Q40
Section 5 | Optimization and Linear Programming | Q41 – Q50
Instructions: Choose the single best answer. Pass: 80% in 90 minutes.
══════════════════════════════════════
SECTION 1: CLASSIFICATION MODELS Q1 – Q10
══════════════════════════════════════
Question 1 of 50
A 34-year-old biostatistician at a pharmaceutical company fits a logistic regression to
predict adverse drug reactions. The model outputs a coefficient of 0.693 for patient age.
After exponentiating the coefficient, she needs to interpret the odds ratio for clinical
staff who are not statistically trained.
A. For every one-year increase in age, the probability of an adverse reaction increases by
69.3%.
B. For every one-year increase in age, the odds of an adverse reaction multiply by
approximately 2.0. ✓ CORRECT
C. For every one-year increase in age, the odds of an adverse reaction increase by 0.693
units.
D. Patients over 65 have exactly twice the probability of an adverse reaction compared
to patients under 65.
Correct Answer: B
,Rationale: Exponentiating a logistic regression coefficient yields the odds ratio, and
e^0.693 is approximately 2.0, meaning the odds multiply by 2 for each one-year
increase. Choice A confuses odds with probability, which are not interchangeable in
logistic regression. In clinical reporting, odds ratios are preferred because they remain
constant across all values of the predictor, unlike probability changes.
Question 2 of 50
A 29-year-old data scientist at a streaming service builds a KNN classifier using 500
user-behavior features to predict subscription cancellations. Despite normalizing all
inputs, the model performs no better than random guessing. Her 52-year-old mentor
suggests the issue might be fundamental to the feature space itself.
A. KNN always fails with normalized data because distance metrics become unstable
after scaling.
B. The model needs a smaller value of k because 500 features require very local
neighborhoods.
C. The problem is that Euclidean distance is undefined in spaces with more than 100
dimensions.
D. In high-dimensional spaces, the distance between any two points tends to become
uniform, making nearest-neighbor concepts less meaningful. ✓ CORRECT
Correct Answer: D
Rationale: The curse of dimensionality causes distances to concentrate and become
nearly identical as the number of features grows, which is exactly what happens with
500 behavioral features. Choice B is incorrect because reducing k does not solve the
fundamental distance concentration problem and often increases variance. This is why
practitioners routinely apply PCA or feature selection before KNN when working with
high-dimensional consumer data.
Question 3 of 50
,A 45-year-old operations researcher at a manufacturing firm is tasked with classifying
defective welds using sensor data. A scatterplot of the two most informative features
reveals a clear circular decision boundary separating good parts from defective ones.
He needs to select an SVM kernel that captures this structure without excessive
complexity.
A. A radial basis function (RBF) kernel would map the data into a higher-dimensional
space where a linear separator can approximate the circular boundary. ✓ CORRECT
B. A linear kernel would suffice if he first squares both features to create polynomial
terms.
C. A sigmoid kernel is the standard choice for circular boundaries because it mimics
logistic regression.
D. A polynomial kernel of degree 20 would guarantee zero training error and generalize
well to new welds.
Correct Answer: A
Rationale: The RBF kernel is specifically designed to handle complex, non-linear
boundaries like circles by implicitly transforming features into an infinite-dimensional
space. Choice D is tempting because high-degree polynomials can fit training data
perfectly, but they almost always overfit and perform poorly on new manufacturing data.
In industrial quality control, model stability is usually more important than training-set
perfection.
Question 4 of 50
A 38-year-old insurance analyst builds a logistic regression with 47 demographic and
driving-history predictors to estimate claim probability. Many variables are highly
correlated, and she wants a model that automatically eliminates irrelevant predictors
while keeping the rest. She is choosing between L1 and L2 regularization.
A. L2 regularization (ridge) would drive many coefficients exactly to zero and produce a
sparse model.
, B. Both L1 and L2 regularization produce identical sparse solutions when the tuning
parameter is large enough.
C. L1 regularization (lasso) tends to shrink some coefficients all the way to zero,
yielding a simpler, interpretable model. ✓ CORRECT
D. Neither L1 nor L2 regularization should be used with logistic regression because
maximum likelihood already prevents overfitting.
Correct Answer: C
Rationale: L1 regularization's geometric diamond-shaped constraint naturally forces
some coefficients to exactly zero, which is the definition of a sparse model. Choice A
incorrectly attributes this property to L2, which shrinks coefficients smoothly toward
zero but rarely eliminates them entirely. In insurance pricing, sparse models are often
preferred by regulators who must audit which factors influence premium calculations.
Question 5 of 50
A 31-year-old fraud analyst at a credit card company deploys a classifier that flags 1% of
transactions as potentially fraudulent. Out of 100,000 transactions, the model correctly
catches 850 fraud cases, misses 150 fraud cases, and incorrectly flags 1,200 legitimate
transactions. The compliance team asks which single metric best captures the model's
ability to find actual fraud without drowning investigators in false alarms.
A. Recall is 85%, which means the model is wrong 15% of the time overall.
B. Precision is approximately 41.5%, indicating that fewer than half of flagged
transactions are actually fraudulent. ✓ CORRECT
C. Accuracy is 98.65%, making this the most reliable metric for the compliance report.
D. Specificity is 98.8%, which proves the model is excellent at detecting fraud in this
population.
Correct Answer: B
Rationale: Precision equals true positives divided by all positive predictions, or 850 /
(850 + 1,200) ≈ 41.5%, which directly answers how many flagged cases are real fraud.
Choice C is a common trap because accuracy is misleadingly high in imbalanced fraud
Exam-Style Questions with Detailed Rationales | 100%
Verified | Pass Guaranteed – A+ Graded
TABLE OF CONTENTS
Section 1 | Classification Models | Q1 – Q10
Section 2 | Clustering and Dimensionality Reduction | Q11 – Q20
Section 3 | Time Series and Forecasting | Q21 – Q30
Section 4 | Validation and Model Selection | Q31 – Q40
Section 5 | Optimization and Linear Programming | Q41 – Q50
Instructions: Choose the single best answer. Pass: 80% in 90 minutes.
══════════════════════════════════════
SECTION 1: CLASSIFICATION MODELS Q1 – Q10
══════════════════════════════════════
Question 1 of 50
A 34-year-old biostatistician at a pharmaceutical company fits a logistic regression to
predict adverse drug reactions. The model outputs a coefficient of 0.693 for patient age.
After exponentiating the coefficient, she needs to interpret the odds ratio for clinical
staff who are not statistically trained.
A. For every one-year increase in age, the probability of an adverse reaction increases by
69.3%.
B. For every one-year increase in age, the odds of an adverse reaction multiply by
approximately 2.0. ✓ CORRECT
C. For every one-year increase in age, the odds of an adverse reaction increase by 0.693
units.
D. Patients over 65 have exactly twice the probability of an adverse reaction compared
to patients under 65.
Correct Answer: B
,Rationale: Exponentiating a logistic regression coefficient yields the odds ratio, and
e^0.693 is approximately 2.0, meaning the odds multiply by 2 for each one-year
increase. Choice A confuses odds with probability, which are not interchangeable in
logistic regression. In clinical reporting, odds ratios are preferred because they remain
constant across all values of the predictor, unlike probability changes.
Question 2 of 50
A 29-year-old data scientist at a streaming service builds a KNN classifier using 500
user-behavior features to predict subscription cancellations. Despite normalizing all
inputs, the model performs no better than random guessing. Her 52-year-old mentor
suggests the issue might be fundamental to the feature space itself.
A. KNN always fails with normalized data because distance metrics become unstable
after scaling.
B. The model needs a smaller value of k because 500 features require very local
neighborhoods.
C. The problem is that Euclidean distance is undefined in spaces with more than 100
dimensions.
D. In high-dimensional spaces, the distance between any two points tends to become
uniform, making nearest-neighbor concepts less meaningful. ✓ CORRECT
Correct Answer: D
Rationale: The curse of dimensionality causes distances to concentrate and become
nearly identical as the number of features grows, which is exactly what happens with
500 behavioral features. Choice B is incorrect because reducing k does not solve the
fundamental distance concentration problem and often increases variance. This is why
practitioners routinely apply PCA or feature selection before KNN when working with
high-dimensional consumer data.
Question 3 of 50
,A 45-year-old operations researcher at a manufacturing firm is tasked with classifying
defective welds using sensor data. A scatterplot of the two most informative features
reveals a clear circular decision boundary separating good parts from defective ones.
He needs to select an SVM kernel that captures this structure without excessive
complexity.
A. A radial basis function (RBF) kernel would map the data into a higher-dimensional
space where a linear separator can approximate the circular boundary. ✓ CORRECT
B. A linear kernel would suffice if he first squares both features to create polynomial
terms.
C. A sigmoid kernel is the standard choice for circular boundaries because it mimics
logistic regression.
D. A polynomial kernel of degree 20 would guarantee zero training error and generalize
well to new welds.
Correct Answer: A
Rationale: The RBF kernel is specifically designed to handle complex, non-linear
boundaries like circles by implicitly transforming features into an infinite-dimensional
space. Choice D is tempting because high-degree polynomials can fit training data
perfectly, but they almost always overfit and perform poorly on new manufacturing data.
In industrial quality control, model stability is usually more important than training-set
perfection.
Question 4 of 50
A 38-year-old insurance analyst builds a logistic regression with 47 demographic and
driving-history predictors to estimate claim probability. Many variables are highly
correlated, and she wants a model that automatically eliminates irrelevant predictors
while keeping the rest. She is choosing between L1 and L2 regularization.
A. L2 regularization (ridge) would drive many coefficients exactly to zero and produce a
sparse model.
, B. Both L1 and L2 regularization produce identical sparse solutions when the tuning
parameter is large enough.
C. L1 regularization (lasso) tends to shrink some coefficients all the way to zero,
yielding a simpler, interpretable model. ✓ CORRECT
D. Neither L1 nor L2 regularization should be used with logistic regression because
maximum likelihood already prevents overfitting.
Correct Answer: C
Rationale: L1 regularization's geometric diamond-shaped constraint naturally forces
some coefficients to exactly zero, which is the definition of a sparse model. Choice A
incorrectly attributes this property to L2, which shrinks coefficients smoothly toward
zero but rarely eliminates them entirely. In insurance pricing, sparse models are often
preferred by regulators who must audit which factors influence premium calculations.
Question 5 of 50
A 31-year-old fraud analyst at a credit card company deploys a classifier that flags 1% of
transactions as potentially fraudulent. Out of 100,000 transactions, the model correctly
catches 850 fraud cases, misses 150 fraud cases, and incorrectly flags 1,200 legitimate
transactions. The compliance team asks which single metric best captures the model's
ability to find actual fraud without drowning investigators in false alarms.
A. Recall is 85%, which means the model is wrong 15% of the time overall.
B. Precision is approximately 41.5%, indicating that fewer than half of flagged
transactions are actually fraudulent. ✓ CORRECT
C. Accuracy is 98.65%, making this the most reliable metric for the compliance report.
D. Specificity is 98.8%, which proves the model is excellent at detecting fraud in this
population.
Correct Answer: B
Rationale: Precision equals true positives divided by all positive predictions, or 850 /
(850 + 1,200) ≈ 41.5%, which directly answers how many flagged cases are real fraud.
Choice C is a common trap because accuracy is misleadingly high in imbalanced fraud
