ISYE 6501 ADVANCED ANALYTICS
MODELING COMPREHENSIVE FINAL
QUESTIONS AND ANSWERS
1. When using Support Vector Machines (SVM), what is the primary purpose of the ‘C’
parameter in the soft-margin formulation?
A. To define the degree of the polynomial kernel.
B. To set the learning rate for the gradient descent optimization.
C. To control the trade-off between maximizing the margin and minimizing classification
errors.
D. To determine the number of support vectors to be used in the final model.
Answer: C
Conceptual Explanation: The C parameter acts as a regularization parameter that
determines the weight given to misclassifications versus the width of the margin. A large C
prioritizes correct classification, while a small C prioritizes a larger margin.
2. In the context of K-means clustering, why is it critical to scale or normalize features before
running the algorithm?
A. K-means is based on Euclidean distance, making it sensitive to the magnitude of feature
values.
,B. Scaling ensures that the algorithm converges in a single iteration.
C. Standardization is required to ensure the covariance matrix is identity.
D. Normalization prevents the centroids from moving outside the data range.
Answer: A
Conceptual Explanation: Since K-means uses distance metrics (typically Euclidean),
features with larger scales will dominate the distance calculation, leading to biased
clusters.
3. Which of the following best describes the ‘Shadow Price’ in a Linear Programming
optimization problem?
A. The cost of implementing the optimal solution.
B. The penalty for violating a non-negativity constraint.
C. The amount by which the objective function value changes per unit increase in a
constraint’s right-hand side.
D. The price at which a secondary market for resources is established.
Answer: C
Conceptual Explanation: The shadow price represents the marginal value of a resource; it
tells you how much the objective (e.g., profit) would improve if a constraint (e.g., labor
hours) were relaxed by one unit.
, 4. In Time Series analysis, what distinguishes a GARCH model from an ARIMA model?
A. ARIMA is a non-parametric model, whereas GARCH is parametric.
B. GARCH is used for seasonal data, while ARIMA is used for trend data.
C. ARIMA models the mean, while GARCH models the variance (volatility).
D. GARCH requires the data to be stationary, while ARIMA does not.
Answer: C
Conceptual Explanation: ARIMA (AutoRegressive Integrated Moving Average) focuses on
predicting the level or mean of the series, while GARCH (Generalized Autoregressive
Conditional Heteroskedasticity) models time-varying variance or volatility.
5. When evaluating a model using the Akaike Information Criterion (AIC), which of the
following is true?
A. A higher AIC value indicates a better-fitting model.
B. AIC only measures the goodness-of-fit without considering model complexity.
C. AIC balances goodness-of-fit with the number of parameters to prevent overfitting.
D. AIC is only applicable to linear regression models.
Answer: C
Conceptual Explanation: AIC rewards goodness of fit (likelihood) but includes a penalty
term for the number of estimated parameters to discourage overly complex models.
MODELING COMPREHENSIVE FINAL
QUESTIONS AND ANSWERS
1. When using Support Vector Machines (SVM), what is the primary purpose of the ‘C’
parameter in the soft-margin formulation?
A. To define the degree of the polynomial kernel.
B. To set the learning rate for the gradient descent optimization.
C. To control the trade-off between maximizing the margin and minimizing classification
errors.
D. To determine the number of support vectors to be used in the final model.
Answer: C
Conceptual Explanation: The C parameter acts as a regularization parameter that
determines the weight given to misclassifications versus the width of the margin. A large C
prioritizes correct classification, while a small C prioritizes a larger margin.
2. In the context of K-means clustering, why is it critical to scale or normalize features before
running the algorithm?
A. K-means is based on Euclidean distance, making it sensitive to the magnitude of feature
values.
,B. Scaling ensures that the algorithm converges in a single iteration.
C. Standardization is required to ensure the covariance matrix is identity.
D. Normalization prevents the centroids from moving outside the data range.
Answer: A
Conceptual Explanation: Since K-means uses distance metrics (typically Euclidean),
features with larger scales will dominate the distance calculation, leading to biased
clusters.
3. Which of the following best describes the ‘Shadow Price’ in a Linear Programming
optimization problem?
A. The cost of implementing the optimal solution.
B. The penalty for violating a non-negativity constraint.
C. The amount by which the objective function value changes per unit increase in a
constraint’s right-hand side.
D. The price at which a secondary market for resources is established.
Answer: C
Conceptual Explanation: The shadow price represents the marginal value of a resource; it
tells you how much the objective (e.g., profit) would improve if a constraint (e.g., labor
hours) were relaxed by one unit.
, 4. In Time Series analysis, what distinguishes a GARCH model from an ARIMA model?
A. ARIMA is a non-parametric model, whereas GARCH is parametric.
B. GARCH is used for seasonal data, while ARIMA is used for trend data.
C. ARIMA models the mean, while GARCH models the variance (volatility).
D. GARCH requires the data to be stationary, while ARIMA does not.
Answer: C
Conceptual Explanation: ARIMA (AutoRegressive Integrated Moving Average) focuses on
predicting the level or mean of the series, while GARCH (Generalized Autoregressive
Conditional Heteroskedasticity) models time-varying variance or volatility.
5. When evaluating a model using the Akaike Information Criterion (AIC), which of the
following is true?
A. A higher AIC value indicates a better-fitting model.
B. AIC only measures the goodness-of-fit without considering model complexity.
C. AIC balances goodness-of-fit with the number of parameters to prevent overfitting.
D. AIC is only applicable to linear regression models.
Answer: C
Conceptual Explanation: AIC rewards goodness of fit (likelihood) but includes a penalty
term for the number of estimated parameters to discourage overly complex models.
