ISYE 6501 FINAL EXAMINATION
ADVANCED ANALYTICS MODELING.
EXAM QUESTIONS AND ANSWERS
1. When using a CUSUM (Cumulative Sum) control chart to detect a change in the mean of a
process, what is the primary purpose of the ‘slack’ variable (often denoted as C or k)?
A. To account for the inherent volatility of the time series data.
B. To increase the sensitivity of the model to very small shifts.
C. To normalize the data points against a standard normal distribution.
D. To prevent the cumulative sum from increasing due to minor, non-significant variations.
Answer: D
Conceptual Explanation: The slack variable represents the distance from the mean that
we are willing to tolerate without signaling a change. It prevents the CUSUM value from
growing due to insignificant noise.
2. In the context of K-Nearest Neighbors (KNN), why is it considered standard practice to scale
or normalize features before calculating distances?
A. To prevent features with larger numerical ranges from dominating the distance
calculation.
,B. To ensure the model converges faster during the optimization phase.
C. To convert categorical data into a continuous numerical format.
D. To reduce the dimensionality of the feature space through linear projection.
Answer: A
Conceptual Explanation: KNN relies on distance metrics (like Euclidean distance). If one
feature has a range of 0-1 and another 0-1000, the latter will disproportionately influence
the distance unless both are scaled.
3. Which of the following best describes the ‘kernel trick’ used in Support Vector Machines
(SVM)?
A. It transforms data into a higher-dimensional space where a linear separator can be
found without explicitly computing the coordinates.
B. It maps data into a lower-dimensional space to speed up computation.
C. It uses a recursive partitioning algorithm to find the optimal split points for
classification.
D. It averages the results of multiple weak learners to reduce variance.
Answer: A
Conceptual Explanation: The kernel trick allows SVMs to operate in a high-dimensional
feature space by computing inner products between the images of pairs of points, making
non-linear classification efficient.
, 4. In time series analysis, what condition must be met for a process to be considered
‘stationary’?
A. The trend component must be linear and strictly increasing over time.
B. The process must exhibit a clear seasonal pattern that repeats every year.
C. The process must have a constant mean, constant variance, and time-independent
autocovariance.
D. The residuals must follow a Poisson distribution.
Answer: C
Conceptual Explanation: Stationarity requires that the statistical properties (mean,
variance, and autocorrelation) of the process do not change over time.
5. Which regularization technique is most appropriate when you suspect only a small subset
of your many predictors are actually relevant to the response variable?
A. Ridge Regression (L2 regularization)
B. Lasso Regression (L1 regularization)
C. Ordinary Least Squares (OLS)
D. Polynomial Regression
Answer: B
Conceptual Explanation: Lasso (L1) has a tendency to set coefficients of less important
variables to exactly zero, effectively performing feature selection.
ADVANCED ANALYTICS MODELING.
EXAM QUESTIONS AND ANSWERS
1. When using a CUSUM (Cumulative Sum) control chart to detect a change in the mean of a
process, what is the primary purpose of the ‘slack’ variable (often denoted as C or k)?
A. To account for the inherent volatility of the time series data.
B. To increase the sensitivity of the model to very small shifts.
C. To normalize the data points against a standard normal distribution.
D. To prevent the cumulative sum from increasing due to minor, non-significant variations.
Answer: D
Conceptual Explanation: The slack variable represents the distance from the mean that
we are willing to tolerate without signaling a change. It prevents the CUSUM value from
growing due to insignificant noise.
2. In the context of K-Nearest Neighbors (KNN), why is it considered standard practice to scale
or normalize features before calculating distances?
A. To prevent features with larger numerical ranges from dominating the distance
calculation.
,B. To ensure the model converges faster during the optimization phase.
C. To convert categorical data into a continuous numerical format.
D. To reduce the dimensionality of the feature space through linear projection.
Answer: A
Conceptual Explanation: KNN relies on distance metrics (like Euclidean distance). If one
feature has a range of 0-1 and another 0-1000, the latter will disproportionately influence
the distance unless both are scaled.
3. Which of the following best describes the ‘kernel trick’ used in Support Vector Machines
(SVM)?
A. It transforms data into a higher-dimensional space where a linear separator can be
found without explicitly computing the coordinates.
B. It maps data into a lower-dimensional space to speed up computation.
C. It uses a recursive partitioning algorithm to find the optimal split points for
classification.
D. It averages the results of multiple weak learners to reduce variance.
Answer: A
Conceptual Explanation: The kernel trick allows SVMs to operate in a high-dimensional
feature space by computing inner products between the images of pairs of points, making
non-linear classification efficient.
, 4. In time series analysis, what condition must be met for a process to be considered
‘stationary’?
A. The trend component must be linear and strictly increasing over time.
B. The process must exhibit a clear seasonal pattern that repeats every year.
C. The process must have a constant mean, constant variance, and time-independent
autocovariance.
D. The residuals must follow a Poisson distribution.
Answer: C
Conceptual Explanation: Stationarity requires that the statistical properties (mean,
variance, and autocorrelation) of the process do not change over time.
5. Which regularization technique is most appropriate when you suspect only a small subset
of your many predictors are actually relevant to the response variable?
A. Ridge Regression (L2 regularization)
B. Lasso Regression (L1 regularization)
C. Ordinary Least Squares (OLS)
D. Polynomial Regression
Answer: B
Conceptual Explanation: Lasso (L1) has a tendency to set coefficients of less important
variables to exactly zero, effectively performing feature selection.
