ISYE 6501 FINAL EXAM - ADVANCED
ANALYTICS MODELING. EXAM
QUESTIONS AND ANSWERS
1. When using a Support Vector Machine (SVM) with a soft margin, what is the primary role
of the constant ‘C’ (the penalty parameter)?
A. It controls the trade-off between maximizing the margin and minimizing classification
errors.
B. It determines the number of support vectors used in the final model.
C. It adjusts the learning rate of the optimization algorithm.
D. It defines the degree of the polynomial kernel used for transformation.
Answer: A
Conceptual Explanation: The parameter C determines the trade-off between the margin
size and the misclassification of training points. A large C focuses on classifying all training
points correctly, while a small C allows for a wider margin and some misclassifications.
2. In the context of the CUSUM algorithm, if we decrease the threshold ‘T’, how does it
typically affect the model’s performance?
A. It increases the time required to detect a significant change.
,B. It decreases the sensitivity of the detector to small shifts.
C. It increases the likelihood of a false positive (false alarm).
D. It has no effect on the false alarm rate.
Answer: C
Conceptual Explanation: Lowering the threshold T makes the system more sensitive,
meaning it will detect changes faster but at the cost of more frequent false alarms.
3. Which of the following methods is specifically designed to handle multicollinearity by
adding a penalty term equivalent to the square of the magnitude of coefficients?
A. Lasso Regression
B. Stepwise Regression
C. Elastic Net
D. Ridge Regression
Answer: D
Conceptual Explanation: Ridge Regression uses L2 regularization, which adds a penalty
proportional to the square of the coefficients. Lasso uses L1 (absolute value), and Elastic
Net uses both.
4. In K-Means clustering, the algorithm is highly sensitive to the initial placement of centroids.
Which technique is commonly used to improve the stability and performance of the results?
A. Normalization is not required for K-Means.
, B. Always setting K equal to the number of features.
C. Avoiding the use of Euclidean distance.
D. Using K-Means++ for initialization.
Answer: D
Conceptual Explanation: K-Means++ is a standard initialization method that spreads out
the initial centroids, leading to faster convergence and more stable clustering results.
5. Which time series model is most appropriate for data that exhibits both a linear trend and
multiplicative seasonality?
A. Simple Exponential Smoothing
B. Double Exponential Smoothing (Holt’s Method)
C. Triple Exponential Smoothing (Holt-Winters)
D. Autoregressive (AR) Model
Answer: C
Conceptual Explanation: Triple Exponential Smoothing (Holt-Winters) accounts for level,
trend, and seasonality. Since the data has both trend and seasonality, this is the most
complete model among the choices.
6. In Principal Component Analysis (PCA), if the first three principal components explain 95%
of the variance, what can be inferred?
A. The original data only had three features.
ANALYTICS MODELING. EXAM
QUESTIONS AND ANSWERS
1. When using a Support Vector Machine (SVM) with a soft margin, what is the primary role
of the constant ‘C’ (the penalty parameter)?
A. It controls the trade-off between maximizing the margin and minimizing classification
errors.
B. It determines the number of support vectors used in the final model.
C. It adjusts the learning rate of the optimization algorithm.
D. It defines the degree of the polynomial kernel used for transformation.
Answer: A
Conceptual Explanation: The parameter C determines the trade-off between the margin
size and the misclassification of training points. A large C focuses on classifying all training
points correctly, while a small C allows for a wider margin and some misclassifications.
2. In the context of the CUSUM algorithm, if we decrease the threshold ‘T’, how does it
typically affect the model’s performance?
A. It increases the time required to detect a significant change.
,B. It decreases the sensitivity of the detector to small shifts.
C. It increases the likelihood of a false positive (false alarm).
D. It has no effect on the false alarm rate.
Answer: C
Conceptual Explanation: Lowering the threshold T makes the system more sensitive,
meaning it will detect changes faster but at the cost of more frequent false alarms.
3. Which of the following methods is specifically designed to handle multicollinearity by
adding a penalty term equivalent to the square of the magnitude of coefficients?
A. Lasso Regression
B. Stepwise Regression
C. Elastic Net
D. Ridge Regression
Answer: D
Conceptual Explanation: Ridge Regression uses L2 regularization, which adds a penalty
proportional to the square of the coefficients. Lasso uses L1 (absolute value), and Elastic
Net uses both.
4. In K-Means clustering, the algorithm is highly sensitive to the initial placement of centroids.
Which technique is commonly used to improve the stability and performance of the results?
A. Normalization is not required for K-Means.
, B. Always setting K equal to the number of features.
C. Avoiding the use of Euclidean distance.
D. Using K-Means++ for initialization.
Answer: D
Conceptual Explanation: K-Means++ is a standard initialization method that spreads out
the initial centroids, leading to faster convergence and more stable clustering results.
5. Which time series model is most appropriate for data that exhibits both a linear trend and
multiplicative seasonality?
A. Simple Exponential Smoothing
B. Double Exponential Smoothing (Holt’s Method)
C. Triple Exponential Smoothing (Holt-Winters)
D. Autoregressive (AR) Model
Answer: C
Conceptual Explanation: Triple Exponential Smoothing (Holt-Winters) accounts for level,
trend, and seasonality. Since the data has both trend and seasonality, this is the most
complete model among the choices.
6. In Principal Component Analysis (PCA), if the first three principal components explain 95%
of the variance, what can be inferred?
A. The original data only had three features.
