ISYE 6501 MIDTERM PRACTICE EXAM– QUESTIONS AND
ANSWERS | VERIFIED AND WELL DETAILED ANSWERS PLUS
RATIONALES | GUARANTEED PASS | LATEST EXAM UPDATE |
EXAM PREP | STUDY GUIDE | PRACTICE TEST| DOWNLOAD
INSTANT PDF
1. A data analyst is building a predictive model for equipment failure and notices that the
training error is near zero, but cross-validation error is exceptionally high. Which
analytical phenomenon is most likely occurring?
A. Underfitting
B. Overfitting
C. Multicollinearity
D. Heteroscedasticity
Overfitting occurs when a model learns the training data too well, including its noise and
random fluctuations, which harms its ability to generalize to unseen data. Underfitting
happens when a model is too simple to capture underlying patterns. Multicollinearity refers to
high correlation among predictor variables, and heteroscedasticity involves non-constant
variance in residuals.
2. Which of the following modeling techniques is considered a non-parametric method for
classification and regression that partitions the feature space into rectangular regions?
A. Linear Regression
B. Logistic Regression
C. Classification and Regression Trees (CART)
D. Linear Discriminant Analysis
CART models make sequential splits on predictor variables to create rectangular regions,
making them non-parametric. Linear regression, logistic regression, and linear discriminant
analysis rely on specific parametric assumptions regarding the underlying data distribution.
3. In regularized linear regression, which penalty term is specifically known for performing
continuous shrinkage and variable selection simultaneously by setting some coefficients
exactly to zero?
A. L1 penalty (Lasso)
B. L2 penalty (Ridge)
C. Elastic Net mixing parameter
D. Maximum likelihood penalty
,The L1 penalty, used in Lasso regression, adds the absolute values of the coefficients to the
loss function, which can drive weak feature coefficients completely to zero for variable
selection. The L2 penalty (Ridge) shrinks coefficients toward zero but does not set them
exactly to zero.
4. When evaluating a classification model predicting a rare medical condition, why might
overall accuracy be a misleading metric?
A. Accuracy penalizes false positives too heavily.
B. A naive model predicting the majority class can achieve extremely high accuracy while
failing to identify any positive cases.
C. Accuracy requires continuous probability outputs rather than discrete classes.
D. Accuracy cannot be calculated when class distributions are skewed.
In highly imbalanced datasets, a classifier that always predicts the majority class will yield a
high accuracy score, masking its complete inability to detect the minority class. Metrics like
precision, recall, and F1-score are better suited for imbalanced data.
5. An industrial engineer wants to forecast weekly manufacturing demand using a time
series model that accounts for both seasonal fluctuations and a long-term upward trend.
Which method is most appropriate?
A. Simple Exponential Smoothing
B. K-Means Clustering
C. Holt-Winters Exponential Smoothing
D. Principal Component Analysis
Holt-Winters exponential smoothing explicitly handles time series data exhibiting both trend
and seasonality. Simple exponential smoothing is only suitable for data without clear trends or
seasonal patterns. K-means and PCA are not time series forecasting techniques.
6. Which validation technique involves partitioning the dataset into $k$ subsets, training
the model on $k-1$ subsets, and testing on the remaining subset, repeating this process $k$
times?
A. Leave-one-out cross-validation
B. K-fold cross-validation
C. Holdout validation
D. Bootstrap resampling
K-fold cross-validation splits the data into $k$ equal parts to iteratively train and validate the
model, providing a robust estimate of out-of-sample performance. Leave-one-out sets $k$
equal to the sample size, while holdout validation splits data only once.
7. In logistic regression, the log-odds of the positive outcome are modeled as which type of
function of the predictor variables?
, A. Quadratic function
B. Exponential function
C. Linear function
D. Logarithmic function
Logistic regression models the natural logarithm of the odds (logit) as a linear combination of
the predictor variables. This ensures that the predicted probabilities remain bounded between
zero and one after applying the logistic function.
8. A retail analyst is grouping store locations based on annual sales volume, customer foot
traffic, and average basket size without any pre-defined labels. Which machine learning
approach should be utilized?
A. Supervised learning
B. Reinforcement learning
C. Unsupervised learning
D. Semi-supervised learning
Unsupervised learning algorithms, such as clustering techniques, analyze unlabeled data to
discover natural groupings or hidden patterns. Supervised learning requires target labels,
which are absent in this customer segmentation scenario.
9. What is the primary purpose of scaling or standardizing numerical features before
applying distance-based algorithms like K-Means clustering?
A. To remove categorical variables from the dataset
B. To ensure that variables with larger numeric ranges do not disproportionately
dominate the distance calculation
C. To eliminate multicollinearity among predictors
D. To convert non-linear relationships into linear ones
Distance-based algorithms calculate Euclidean distance across all features. If one feature has
a much larger scale than others, it will skew the distance metric, making feature scaling
essential.
10. When designing a time series model, what does an autoregressive (AR) term of order
$p$ signify?
A. The model uses past error terms up to lag $q$ to predict future values.
B. The model uses past values of the target variable up to lag $p$ to predict future values.
C. The model includes a seasonal differencing parameter of order $p$.
D. The model incorporates exogenous economic indicators.
An autoregressive model of order $p$ regresses the current value of the series on its own
previous values spanning $p$ time lags. Moving average (MA) terms utilize past forecast
errors instead.
ANSWERS | VERIFIED AND WELL DETAILED ANSWERS PLUS
RATIONALES | GUARANTEED PASS | LATEST EXAM UPDATE |
EXAM PREP | STUDY GUIDE | PRACTICE TEST| DOWNLOAD
INSTANT PDF
1. A data analyst is building a predictive model for equipment failure and notices that the
training error is near zero, but cross-validation error is exceptionally high. Which
analytical phenomenon is most likely occurring?
A. Underfitting
B. Overfitting
C. Multicollinearity
D. Heteroscedasticity
Overfitting occurs when a model learns the training data too well, including its noise and
random fluctuations, which harms its ability to generalize to unseen data. Underfitting
happens when a model is too simple to capture underlying patterns. Multicollinearity refers to
high correlation among predictor variables, and heteroscedasticity involves non-constant
variance in residuals.
2. Which of the following modeling techniques is considered a non-parametric method for
classification and regression that partitions the feature space into rectangular regions?
A. Linear Regression
B. Logistic Regression
C. Classification and Regression Trees (CART)
D. Linear Discriminant Analysis
CART models make sequential splits on predictor variables to create rectangular regions,
making them non-parametric. Linear regression, logistic regression, and linear discriminant
analysis rely on specific parametric assumptions regarding the underlying data distribution.
3. In regularized linear regression, which penalty term is specifically known for performing
continuous shrinkage and variable selection simultaneously by setting some coefficients
exactly to zero?
A. L1 penalty (Lasso)
B. L2 penalty (Ridge)
C. Elastic Net mixing parameter
D. Maximum likelihood penalty
,The L1 penalty, used in Lasso regression, adds the absolute values of the coefficients to the
loss function, which can drive weak feature coefficients completely to zero for variable
selection. The L2 penalty (Ridge) shrinks coefficients toward zero but does not set them
exactly to zero.
4. When evaluating a classification model predicting a rare medical condition, why might
overall accuracy be a misleading metric?
A. Accuracy penalizes false positives too heavily.
B. A naive model predicting the majority class can achieve extremely high accuracy while
failing to identify any positive cases.
C. Accuracy requires continuous probability outputs rather than discrete classes.
D. Accuracy cannot be calculated when class distributions are skewed.
In highly imbalanced datasets, a classifier that always predicts the majority class will yield a
high accuracy score, masking its complete inability to detect the minority class. Metrics like
precision, recall, and F1-score are better suited for imbalanced data.
5. An industrial engineer wants to forecast weekly manufacturing demand using a time
series model that accounts for both seasonal fluctuations and a long-term upward trend.
Which method is most appropriate?
A. Simple Exponential Smoothing
B. K-Means Clustering
C. Holt-Winters Exponential Smoothing
D. Principal Component Analysis
Holt-Winters exponential smoothing explicitly handles time series data exhibiting both trend
and seasonality. Simple exponential smoothing is only suitable for data without clear trends or
seasonal patterns. K-means and PCA are not time series forecasting techniques.
6. Which validation technique involves partitioning the dataset into $k$ subsets, training
the model on $k-1$ subsets, and testing on the remaining subset, repeating this process $k$
times?
A. Leave-one-out cross-validation
B. K-fold cross-validation
C. Holdout validation
D. Bootstrap resampling
K-fold cross-validation splits the data into $k$ equal parts to iteratively train and validate the
model, providing a robust estimate of out-of-sample performance. Leave-one-out sets $k$
equal to the sample size, while holdout validation splits data only once.
7. In logistic regression, the log-odds of the positive outcome are modeled as which type of
function of the predictor variables?
, A. Quadratic function
B. Exponential function
C. Linear function
D. Logarithmic function
Logistic regression models the natural logarithm of the odds (logit) as a linear combination of
the predictor variables. This ensures that the predicted probabilities remain bounded between
zero and one after applying the logistic function.
8. A retail analyst is grouping store locations based on annual sales volume, customer foot
traffic, and average basket size without any pre-defined labels. Which machine learning
approach should be utilized?
A. Supervised learning
B. Reinforcement learning
C. Unsupervised learning
D. Semi-supervised learning
Unsupervised learning algorithms, such as clustering techniques, analyze unlabeled data to
discover natural groupings or hidden patterns. Supervised learning requires target labels,
which are absent in this customer segmentation scenario.
9. What is the primary purpose of scaling or standardizing numerical features before
applying distance-based algorithms like K-Means clustering?
A. To remove categorical variables from the dataset
B. To ensure that variables with larger numeric ranges do not disproportionately
dominate the distance calculation
C. To eliminate multicollinearity among predictors
D. To convert non-linear relationships into linear ones
Distance-based algorithms calculate Euclidean distance across all features. If one feature has
a much larger scale than others, it will skew the distance metric, making feature scaling
essential.
10. When designing a time series model, what does an autoregressive (AR) term of order
$p$ signify?
A. The model uses past error terms up to lag $q$ to predict future values.
B. The model uses past values of the target variable up to lag $p$ to predict future values.
C. The model includes a seasonal differencing parameter of order $p$.
D. The model incorporates exogenous economic indicators.
An autoregressive model of order $p$ regresses the current value of the series on its own
previous values spanning $p$ time lags. Moving average (MA) terms utilize past forecast
errors instead.
