ISYE6501x Intro to Analytics Modeling Final Quiz Summer
2026/2027 Actual Exam – Complete Exam-Style Questions
with Detailed Rationales | 100% Verified | Pass Guaranteed –
A+ Graded
SECTION A: MODEL SELECTION & PROBLEM TYPES (Questions 1–15)
Q1: A retail company has collected 50 features describing customer demographics,
purchase history, and website behavior. They want to identify which subset of features
is most predictive of whether a customer will churn in the next 30 days. Which analytical
approach is best suited for this task?
A. k-means clustering
B. ARIMA modeling
C. Lasso regression for variable selection and/or prediction from feature data
[CORRECT]
D. Exponential smoothing
Correct Answer: C
Rationale: Lasso regression (L1 regularization) performs variable selection by shrinking
some coefficients to exactly zero, making it ideal for identifying the most predictive
features among many candidates. k-means is for unsupervised clustering, ARIMA is for
time-series forecasting, and exponential smoothing is for time-series prediction, not
feature selection.
,Q2: An energy utility needs to forecast daily electricity demand for the next 90 days
using the past 5 years of hourly demand data that exhibits strong daily and weekly
seasonal patterns. Which model is most appropriate for this prediction task?
A. Logistic regression
B. Support Vector Machine with RBF kernel
C. Holt-Winters exponential smoothing for trend and seasonality [CORRECT]
D. k-nearest neighbors
Correct Answer: C
Rationale: Holt-Winters exponential smoothing extends simple exponential smoothing
to capture both trend and seasonality—exactly what electricity demand data exhibits.
Logistic regression and SVM are classification/feature prediction models, and k-NN is
not designed for time-series with trend and seasonality.
Q3: A pharmaceutical company is running a clinical trial to study the effects of 6
different factors (dosage, timing, patient age group, formulation, co-medication, and
diet) on drug efficacy. Testing all possible combinations would require 64 experiments,
which is too expensive. Which approach allows them to study the main effects
efficiently with fewer experiments?
A. k-means clustering
B. Fractional factorial design [CORRECT]
C. Principal Component Analysis
D. GARCH modeling
,Correct Answer: B
Rationale: Fractional factorial design is an experimental design technique that studies
the effects of multiple factors by testing only a carefully selected subset of factor
combinations, dramatically reducing the number of required experiments while still
estimating main effects. k-means and PCA are unsupervised methods, and GARCH
models time-series volatility.
Q4: A financial analyst needs to model the volatility of daily stock returns, where the
volatility itself changes over time (periods of high volatility cluster together). Which
model is specifically designed for this type of time-series behavior?
A. Linear regression
B. GARCH (Generalized Autoregressive Conditional Heteroskedasticity) [CORRECT]
C. k-means clustering
D. Mann-Whitney test
Correct Answer: B
Rationale: GARCH models are specifically designed for time-series data where volatility
(variance) is not constant but changes over time, with clustering of high and low
volatility periods. Linear regression assumes constant variance, k-means is for
clustering, and Mann-Whitney is a nonparametric test for comparing two independent
samples.
Q5: A hospital wants to predict whether a patient will be readmitted within 30 days of
discharge based on 40 clinical and demographic features. Which of the following
models is best suited for this binary classification task using feature data?
, A. ARIMA
B. Logistic regression for classification and prediction from feature data [CORRECT]
C. Exponential smoothing
D. Markov chains
Correct Answer: B
Rationale: Logistic regression is the canonical model for binary classification using
feature data, estimating the probability of readmission as a function of the input
features. ARIMA and exponential smoothing are time-series models, and Markov chains
model state transitions in sequences, not feature-based binary classification.
Q6: A manufacturing quality engineer needs to compare the defect rates of two different
production lines using paired samples, where each sample is recorded as defective or
non-defective (yes/no). Which statistical test is most appropriate?
A. Mann-Whitney test
B. McNemar's test for comparing paired samples with yes/no output [CORRECT]
C. Linear regression
D. t-test for independent samples
Correct Answer: B
Rationale: McNemar's test is specifically designed for comparing paired categorical
data with binary (yes/no) outcomes, such as before/after or paired production line
measurements. Mann-Whitney is for independent samples, linear regression requires
2026/2027 Actual Exam – Complete Exam-Style Questions
with Detailed Rationales | 100% Verified | Pass Guaranteed –
A+ Graded
SECTION A: MODEL SELECTION & PROBLEM TYPES (Questions 1–15)
Q1: A retail company has collected 50 features describing customer demographics,
purchase history, and website behavior. They want to identify which subset of features
is most predictive of whether a customer will churn in the next 30 days. Which analytical
approach is best suited for this task?
A. k-means clustering
B. ARIMA modeling
C. Lasso regression for variable selection and/or prediction from feature data
[CORRECT]
D. Exponential smoothing
Correct Answer: C
Rationale: Lasso regression (L1 regularization) performs variable selection by shrinking
some coefficients to exactly zero, making it ideal for identifying the most predictive
features among many candidates. k-means is for unsupervised clustering, ARIMA is for
time-series forecasting, and exponential smoothing is for time-series prediction, not
feature selection.
,Q2: An energy utility needs to forecast daily electricity demand for the next 90 days
using the past 5 years of hourly demand data that exhibits strong daily and weekly
seasonal patterns. Which model is most appropriate for this prediction task?
A. Logistic regression
B. Support Vector Machine with RBF kernel
C. Holt-Winters exponential smoothing for trend and seasonality [CORRECT]
D. k-nearest neighbors
Correct Answer: C
Rationale: Holt-Winters exponential smoothing extends simple exponential smoothing
to capture both trend and seasonality—exactly what electricity demand data exhibits.
Logistic regression and SVM are classification/feature prediction models, and k-NN is
not designed for time-series with trend and seasonality.
Q3: A pharmaceutical company is running a clinical trial to study the effects of 6
different factors (dosage, timing, patient age group, formulation, co-medication, and
diet) on drug efficacy. Testing all possible combinations would require 64 experiments,
which is too expensive. Which approach allows them to study the main effects
efficiently with fewer experiments?
A. k-means clustering
B. Fractional factorial design [CORRECT]
C. Principal Component Analysis
D. GARCH modeling
,Correct Answer: B
Rationale: Fractional factorial design is an experimental design technique that studies
the effects of multiple factors by testing only a carefully selected subset of factor
combinations, dramatically reducing the number of required experiments while still
estimating main effects. k-means and PCA are unsupervised methods, and GARCH
models time-series volatility.
Q4: A financial analyst needs to model the volatility of daily stock returns, where the
volatility itself changes over time (periods of high volatility cluster together). Which
model is specifically designed for this type of time-series behavior?
A. Linear regression
B. GARCH (Generalized Autoregressive Conditional Heteroskedasticity) [CORRECT]
C. k-means clustering
D. Mann-Whitney test
Correct Answer: B
Rationale: GARCH models are specifically designed for time-series data where volatility
(variance) is not constant but changes over time, with clustering of high and low
volatility periods. Linear regression assumes constant variance, k-means is for
clustering, and Mann-Whitney is a nonparametric test for comparing two independent
samples.
Q5: A hospital wants to predict whether a patient will be readmitted within 30 days of
discharge based on 40 clinical and demographic features. Which of the following
models is best suited for this binary classification task using feature data?
, A. ARIMA
B. Logistic regression for classification and prediction from feature data [CORRECT]
C. Exponential smoothing
D. Markov chains
Correct Answer: B
Rationale: Logistic regression is the canonical model for binary classification using
feature data, estimating the probability of readmission as a function of the input
features. ARIMA and exponential smoothing are time-series models, and Markov chains
model state transitions in sequences, not feature-based binary classification.
Q6: A manufacturing quality engineer needs to compare the defect rates of two different
production lines using paired samples, where each sample is recorded as defective or
non-defective (yes/no). Which statistical test is most appropriate?
A. Mann-Whitney test
B. McNemar's test for comparing paired samples with yes/no output [CORRECT]
C. Linear regression
D. t-test for independent samples
Correct Answer: B
Rationale: McNemar's test is specifically designed for comparing paired categorical
data with binary (yes/no) outcomes, such as before/after or paired production line
measurements. Mann-Whitney is for independent samples, linear regression requires
