ISYE-6501 Intro to Analytics Modeling
Complete Exam Questions and Answers
Questions and Solutions
Question 1
In Support Vector Machines (SVM), what is the role of the soft margin parameter C?
A. It determines the number of clusters in the dataset.
B. It sets the learning rate for gradient descent optimization.
C. It controls the trade-off between maximizing the margin and minimizing classification errors.
D. It specifies the degree of the polynomial kernel used.
Correct Answer: C
Rationale: The soft margin parameter C in SVM balances the dual objectives of maximizing the
margin width and penalizing misclassified points or points within the margin. A larger C places
more emphasis on classifying training points correctly, whereas a smaller C encourages a wider
margin at the cost of some classification errors.
Question 2
When applying K-means clustering, which of the following techniques is commonly used to help
select an appropriate number of clusters, k?
A. Receiver Operating Characteristic (ROC) curve
B. Elbow method using total within-cluster sum of squares
C. Durbin-Watson statistic
D. Akaike Information Criterion (AIC)
Correct Answer: B
Rationale: The elbow method plots the total within-cluster sum of squares (WSS) against
varying values of k. The point where the rate of decrease abruptly changes ("elbow") indicates a
suitable balance between variance explained and cluster complexity.
Question 3
What is the primary objective of CUSUM (Cumulative Sum) quality control charts?
A. To detect small, persistent shifts in the mean of a process over time.
B. To impute missing values in high-dimensional time series data.
C. To perform cross-validation on time-series regression models.
D. To cluster temporal data into non-overlapping seasonal regimes.
Correct Answer: A
Rationale: CUSUM accumulates deviations from a target baseline value over time, making it
particularly sensitive and effective for detecting small, sustained shifts in a process mean that
standard Shewhart control charts might miss.
Question 4
In Exponential Smoothing (Holt-Winters), which parameter specifically handles the trend
,component of the time series?
A. Alpha (α)
B. Beta (β)
C. Gamma (γ)
D. Delta (δ)
Correct Answer: B
Rationale: In Holt-Winters exponential smoothing, alpha (α) controls the smoothing of the base
level, beta (β) controls the smoothing of the trend component, and gamma (γ) handles seasonal
variations.
Question 5
Which feature selection technique penalizes the absolute value of regression coefficients,
effectively driving some coefficients to exactly zero?
A. Ridge Regression (L2 regularization)
B. Principal Component Analysis (PCA)
C. Lasso Regression (L1 regularization)
D. Ordinary Least Squares (OLS)
Correct Answer: C
Rationale: Lasso regression adds an L1 penalty equal to the absolute magnitude of the
coefficients. Because of the geometry of L1 constraints, it forces coefficients of non-essential
features to zero, performing automatic feature selection.
Question 6
When using Principal Component Analysis (PCA) for dimension reduction, how are principal
components constructed?
A. As non-linear kernel transformations that maximize cluster density.
B. As random projections of original variables into lower-dimensional space.
C. As linear combinations of original features that are uncorrelated and sequentially maximize
variance.
D. As decision trees that split variables based on entropy reduction.
Correct Answer: C
Rationale: PCA creates orthogonal (uncorrelated) linear combinations of original predictors.
The first principal component accounts for the maximum variance, and each subsequent
component accounts for the largest remaining variance under the orthogonality constraint.
Question 7
In a p-value evaluation for hypothesis testing in linear regression, what does a p-value of 0.02
typically indicate at a significance level of α = 0.05?
A. There is insufficient evidence to reject the null hypothesis.
B. The model explains 98% of the total variance in the outcome variable.
C. Reject the null hypothesis, concluding the feature has a statistically significant relationship
with the response.
D. The probability that the alternative hypothesis is false is exactly 2%.
Correct Answer: C
Rationale: Since the p-value (0.02) is less than the standard significance threshold of 0.05, we
reject the null hypothesis, indicating that the predictor's observed effect is statistically significant.
, Question 8
Which method is designed to address multicollinearity among predictor variables in a multiple
linear regression model?
A. Ridge Regression
B. K-Nearest Neighbors Classification
C. Grubbs' Test
D. Chi-Square Test of Independence
Correct Answer: A
Rationale: Ridge regression adds an L2 penalty term to the sum of squared errors, which
stabilizes coefficient estimates when predictors are highly correlated (multicollinear), reducing
model variance.
Question 9
In logistic regression, what transformation is applied to express the probability p as a linear
combination of predictors?
A. Box-Cox transformation
B. Logit transformation, ln(p / (1 - p))
C. Logarithmic variance transformation
D. Min-Max normalization
Correct Answer: B
Rationale: Logistic regression uses the logit function (log-odds), defined as ln(p / (1 - p)), to
map probabilities bounded in [0,1] to an unbounded continuous range suitable for linear
modeling.
Question 10
What is the primary purpose of cross-validation (e.g., k-fold cross-validation) during model
building?
A. To eliminate the need for training data when fitting parameters.
B. To estimate out-of-sample performance and prevent overfitting.
C. To convert continuous predictors into categorical features.
D. To guarantee that regression residuals follow a normal distribution.
Correct Answer: B
Rationale: K-fold cross-validation partitions data into k subsets, training on k-1 folds and
validating on the remaining fold. Averaging performance across all folds provides an unbiased
estimate of how well the model generalizes to unseen data.
Question 11
What is a key difference between Discrete Event Simulation (DES) and System Dynamics
modeling?
A. DES tracks continuous differential equations, while System Dynamics tracks discrete objects.
B. DES models systems as sequences of individual events in time, whereas System Dynamics
focuses on feedback loops, stocks, and flows.
C. DES can only be applied to financial portfolios, whereas System Dynamics is strictly for
supply chains.
D. DES requires deterministic parameters, while System Dynamics requires stochastic
variables.
Complete Exam Questions and Answers
Questions and Solutions
Question 1
In Support Vector Machines (SVM), what is the role of the soft margin parameter C?
A. It determines the number of clusters in the dataset.
B. It sets the learning rate for gradient descent optimization.
C. It controls the trade-off between maximizing the margin and minimizing classification errors.
D. It specifies the degree of the polynomial kernel used.
Correct Answer: C
Rationale: The soft margin parameter C in SVM balances the dual objectives of maximizing the
margin width and penalizing misclassified points or points within the margin. A larger C places
more emphasis on classifying training points correctly, whereas a smaller C encourages a wider
margin at the cost of some classification errors.
Question 2
When applying K-means clustering, which of the following techniques is commonly used to help
select an appropriate number of clusters, k?
A. Receiver Operating Characteristic (ROC) curve
B. Elbow method using total within-cluster sum of squares
C. Durbin-Watson statistic
D. Akaike Information Criterion (AIC)
Correct Answer: B
Rationale: The elbow method plots the total within-cluster sum of squares (WSS) against
varying values of k. The point where the rate of decrease abruptly changes ("elbow") indicates a
suitable balance between variance explained and cluster complexity.
Question 3
What is the primary objective of CUSUM (Cumulative Sum) quality control charts?
A. To detect small, persistent shifts in the mean of a process over time.
B. To impute missing values in high-dimensional time series data.
C. To perform cross-validation on time-series regression models.
D. To cluster temporal data into non-overlapping seasonal regimes.
Correct Answer: A
Rationale: CUSUM accumulates deviations from a target baseline value over time, making it
particularly sensitive and effective for detecting small, sustained shifts in a process mean that
standard Shewhart control charts might miss.
Question 4
In Exponential Smoothing (Holt-Winters), which parameter specifically handles the trend
,component of the time series?
A. Alpha (α)
B. Beta (β)
C. Gamma (γ)
D. Delta (δ)
Correct Answer: B
Rationale: In Holt-Winters exponential smoothing, alpha (α) controls the smoothing of the base
level, beta (β) controls the smoothing of the trend component, and gamma (γ) handles seasonal
variations.
Question 5
Which feature selection technique penalizes the absolute value of regression coefficients,
effectively driving some coefficients to exactly zero?
A. Ridge Regression (L2 regularization)
B. Principal Component Analysis (PCA)
C. Lasso Regression (L1 regularization)
D. Ordinary Least Squares (OLS)
Correct Answer: C
Rationale: Lasso regression adds an L1 penalty equal to the absolute magnitude of the
coefficients. Because of the geometry of L1 constraints, it forces coefficients of non-essential
features to zero, performing automatic feature selection.
Question 6
When using Principal Component Analysis (PCA) for dimension reduction, how are principal
components constructed?
A. As non-linear kernel transformations that maximize cluster density.
B. As random projections of original variables into lower-dimensional space.
C. As linear combinations of original features that are uncorrelated and sequentially maximize
variance.
D. As decision trees that split variables based on entropy reduction.
Correct Answer: C
Rationale: PCA creates orthogonal (uncorrelated) linear combinations of original predictors.
The first principal component accounts for the maximum variance, and each subsequent
component accounts for the largest remaining variance under the orthogonality constraint.
Question 7
In a p-value evaluation for hypothesis testing in linear regression, what does a p-value of 0.02
typically indicate at a significance level of α = 0.05?
A. There is insufficient evidence to reject the null hypothesis.
B. The model explains 98% of the total variance in the outcome variable.
C. Reject the null hypothesis, concluding the feature has a statistically significant relationship
with the response.
D. The probability that the alternative hypothesis is false is exactly 2%.
Correct Answer: C
Rationale: Since the p-value (0.02) is less than the standard significance threshold of 0.05, we
reject the null hypothesis, indicating that the predictor's observed effect is statistically significant.
, Question 8
Which method is designed to address multicollinearity among predictor variables in a multiple
linear regression model?
A. Ridge Regression
B. K-Nearest Neighbors Classification
C. Grubbs' Test
D. Chi-Square Test of Independence
Correct Answer: A
Rationale: Ridge regression adds an L2 penalty term to the sum of squared errors, which
stabilizes coefficient estimates when predictors are highly correlated (multicollinear), reducing
model variance.
Question 9
In logistic regression, what transformation is applied to express the probability p as a linear
combination of predictors?
A. Box-Cox transformation
B. Logit transformation, ln(p / (1 - p))
C. Logarithmic variance transformation
D. Min-Max normalization
Correct Answer: B
Rationale: Logistic regression uses the logit function (log-odds), defined as ln(p / (1 - p)), to
map probabilities bounded in [0,1] to an unbounded continuous range suitable for linear
modeling.
Question 10
What is the primary purpose of cross-validation (e.g., k-fold cross-validation) during model
building?
A. To eliminate the need for training data when fitting parameters.
B. To estimate out-of-sample performance and prevent overfitting.
C. To convert continuous predictors into categorical features.
D. To guarantee that regression residuals follow a normal distribution.
Correct Answer: B
Rationale: K-fold cross-validation partitions data into k subsets, training on k-1 folds and
validating on the remaining fold. Averaging performance across all folds provides an unbiased
estimate of how well the model generalizes to unseen data.
Question 11
What is a key difference between Discrete Event Simulation (DES) and System Dynamics
modeling?
A. DES tracks continuous differential equations, while System Dynamics tracks discrete objects.
B. DES models systems as sequences of individual events in time, whereas System Dynamics
focuses on feedback loops, stocks, and flows.
C. DES can only be applied to financial portfolios, whereas System Dynamics is strictly for
supply chains.
D. DES requires deterministic parameters, while System Dynamics requires stochastic
variables.
