ISYE 6501 INTRODUCTION TO
ANALYTICS MODELING FINAL EXAM
PREP QUESTIONS AND ANSWERS
1. In Support Vector Machines (SVM), what is the effect of choosing a very large value for the
cost parameter ‘C’?
A. It creates a harder margin, potentially leading to overfitting by minimizing training
errors strictly.
B. It leads to a wider margin and more misclassified points in the training set.
C. It increases the number of support vectors significantly, simplifying the model.
D. It forces the model to ignore outliers by maximizing the margin width.
Answer: A
Conceptual Explanation: A large ‘C’ value penalizes misclassifications heavily, leading to a
narrower (harder) margin that tries to classify all training points correctly, which can
cause overfitting.
,2. Which time series model is specifically designed to handle volatility clustering, where
periods of high variance are followed by high variance?
A. ARIMA
B. GARCH
C. Holt-Winters
D. Simple Exponential Smoothing
Answer: B
Conceptual Explanation: GARCH (Generalized Autoregressive Conditional
Heteroskedasticity) is designed to model and forecast time-varying volatility.
3. When using K-fold cross-validation, why do we typically set aside a completely separate
test set before starting the process?
A. To calculate the training error more efficiently.
B. To ensure the model has enough data for the validation folds.
C. To get an unbiased estimate of model performance on truly unseen data after
hyperparameter tuning.
D. To reduce the computational complexity of the cross-validation algorithm.
Answer: C
, Conceptual Explanation: Cross-validation is used for model selection and tuning; a
separate test set is required for the final evaluation to avoid ‘data leakage’ from the tuning
process.
4. In the context of CUSUM (Cumulative Sum Control Chart), what does the parameter ‘k’
(slack value) represent?
A. The threshold for triggering an alarm.
B. The standard deviation of the underlying process.
C. The amount of shift in the mean we are willing to tolerate without signaling.
D. The number of observations to average before plotting.
Answer: C
Conceptual Explanation: The slack value ‘k’ represents the allowance for half the shift in
the mean that we want to detect; it helps prevent small fluctuations from triggering false
alarms.
5. Which of the following methods uses L1 regularization to perform simultaneous feature
selection and shrinkage?
A. Ridge Regression
B. Ordinary Least Squares
C. Elastic Net
D. Lasso Regression
ANALYTICS MODELING FINAL EXAM
PREP QUESTIONS AND ANSWERS
1. In Support Vector Machines (SVM), what is the effect of choosing a very large value for the
cost parameter ‘C’?
A. It creates a harder margin, potentially leading to overfitting by minimizing training
errors strictly.
B. It leads to a wider margin and more misclassified points in the training set.
C. It increases the number of support vectors significantly, simplifying the model.
D. It forces the model to ignore outliers by maximizing the margin width.
Answer: A
Conceptual Explanation: A large ‘C’ value penalizes misclassifications heavily, leading to a
narrower (harder) margin that tries to classify all training points correctly, which can
cause overfitting.
,2. Which time series model is specifically designed to handle volatility clustering, where
periods of high variance are followed by high variance?
A. ARIMA
B. GARCH
C. Holt-Winters
D. Simple Exponential Smoothing
Answer: B
Conceptual Explanation: GARCH (Generalized Autoregressive Conditional
Heteroskedasticity) is designed to model and forecast time-varying volatility.
3. When using K-fold cross-validation, why do we typically set aside a completely separate
test set before starting the process?
A. To calculate the training error more efficiently.
B. To ensure the model has enough data for the validation folds.
C. To get an unbiased estimate of model performance on truly unseen data after
hyperparameter tuning.
D. To reduce the computational complexity of the cross-validation algorithm.
Answer: C
, Conceptual Explanation: Cross-validation is used for model selection and tuning; a
separate test set is required for the final evaluation to avoid ‘data leakage’ from the tuning
process.
4. In the context of CUSUM (Cumulative Sum Control Chart), what does the parameter ‘k’
(slack value) represent?
A. The threshold for triggering an alarm.
B. The standard deviation of the underlying process.
C. The amount of shift in the mean we are willing to tolerate without signaling.
D. The number of observations to average before plotting.
Answer: C
Conceptual Explanation: The slack value ‘k’ represents the allowance for half the shift in
the mean that we want to detect; it helps prevent small fluctuations from triggering false
alarms.
5. Which of the following methods uses L1 regularization to perform simultaneous feature
selection and shrinkage?
A. Ridge Regression
B. Ordinary Least Squares
C. Elastic Net
D. Lasso Regression
