BUS3060 Business Analytics
Midterm Exam Review (Qns & Ans)
2025
Question 1:
Case Study: A retail company develops a multiple regression
model to predict customer lifetime value (CLV) using historical
data. The initial model produces an R‑squared of 0.75. After
adding an additional predictor—which turns out to be irrelevant—
the R‑squared increases to 0.77, but the adjusted R‑squared falls
to 0.74.
Question: What does this outcome most likely indicate?
A. The new predictor substantially improves the model’s
predictive power.
B. The increase in R‑squared is spurious; the model may be
overfitting and the new predictor is not useful.
C. The model now perfectly predicts CLV.
©2025
,D. The adjusted R‑squared always increases when more
predictors are added.
Correct ANS: B. The increase in R‑squared is spurious; the
model may be overfitting and the new predictor is not useful.
Rationale: While R‑squared always increases (or remains the
same) with additional predictors, the adjusted R‑squared penalizes
for added variables. A decrease in adjusted R‑squared suggests
that the new predictor does not add explanatory power and may
even harm the model’s generalizability.
---
Question 2:
Case Study: A company uses a time‑series model to forecast
monthly revenue. The historical data exhibits an upward trend and
strong seasonal fluctuations. The analyst opts for a model that
incorporates both trend and seasonal components.
Question: Which forecasting method is most appropriate for
this scenario?
A. Simple Exponential Smoothing
B. Holt’s Linear Trend Method
C. Holt‑Winters Seasonal Method
D. Moving Average
©2025
, Correct ANS: C. Holt‑Winters Seasonal Method
Rationale: The Holt‑Winters method extends exponential
smoothing to capture both trend and seasonality, making it ideal
for time‑series data with regular seasonal patterns.
---
Question 3:
Case Study: A business uses a decision tree classifier to segment
potential customers based on likelihood to convert. However, the
tree model appears overly complex and is overfitting the training
data.
Question: Which technique is most effective for reducing
overfitting in this decision tree?
A. Increasing the maximum tree depth
B. Pruning the tree
C. Removing cross‑validation steps
D. Adding more independent variables
Correct ANS: B. Pruning the tree
©2025
, Rationale: Pruning is used to remove branches that offer little
power in predicting the target variable, thereby reducing model
complexity and helping to avoid overfitting.
---
Question 4:
Case Study: A marketing analyst uses K‑means clustering to
segment customers. To determine the optimal number of clusters,
the analyst evaluates the within‑cluster sum of squares (WCSS)
for different values of k.
Question: Which method is best for selecting the appropriate
number of clusters?
A. The Elbow Method
B. Linear Discriminant Analysis
C. Principal Component Analysis
D. Hierarchical Clustering
Correct ANS: A. The Elbow Method
Rationale: The Elbow Method involves plotting the WCSS
against the number of clusters and selecting the k at which the
marginal decrease sharply levels off (“the elbow”), indicating an
optimal trade‑off between fit and complexity.
©2025
Midterm Exam Review (Qns & Ans)
2025
Question 1:
Case Study: A retail company develops a multiple regression
model to predict customer lifetime value (CLV) using historical
data. The initial model produces an R‑squared of 0.75. After
adding an additional predictor—which turns out to be irrelevant—
the R‑squared increases to 0.77, but the adjusted R‑squared falls
to 0.74.
Question: What does this outcome most likely indicate?
A. The new predictor substantially improves the model’s
predictive power.
B. The increase in R‑squared is spurious; the model may be
overfitting and the new predictor is not useful.
C. The model now perfectly predicts CLV.
©2025
,D. The adjusted R‑squared always increases when more
predictors are added.
Correct ANS: B. The increase in R‑squared is spurious; the
model may be overfitting and the new predictor is not useful.
Rationale: While R‑squared always increases (or remains the
same) with additional predictors, the adjusted R‑squared penalizes
for added variables. A decrease in adjusted R‑squared suggests
that the new predictor does not add explanatory power and may
even harm the model’s generalizability.
---
Question 2:
Case Study: A company uses a time‑series model to forecast
monthly revenue. The historical data exhibits an upward trend and
strong seasonal fluctuations. The analyst opts for a model that
incorporates both trend and seasonal components.
Question: Which forecasting method is most appropriate for
this scenario?
A. Simple Exponential Smoothing
B. Holt’s Linear Trend Method
C. Holt‑Winters Seasonal Method
D. Moving Average
©2025
, Correct ANS: C. Holt‑Winters Seasonal Method
Rationale: The Holt‑Winters method extends exponential
smoothing to capture both trend and seasonality, making it ideal
for time‑series data with regular seasonal patterns.
---
Question 3:
Case Study: A business uses a decision tree classifier to segment
potential customers based on likelihood to convert. However, the
tree model appears overly complex and is overfitting the training
data.
Question: Which technique is most effective for reducing
overfitting in this decision tree?
A. Increasing the maximum tree depth
B. Pruning the tree
C. Removing cross‑validation steps
D. Adding more independent variables
Correct ANS: B. Pruning the tree
©2025
, Rationale: Pruning is used to remove branches that offer little
power in predicting the target variable, thereby reducing model
complexity and helping to avoid overfitting.
---
Question 4:
Case Study: A marketing analyst uses K‑means clustering to
segment customers. To determine the optimal number of clusters,
the analyst evaluates the within‑cluster sum of squares (WCSS)
for different values of k.
Question: Which method is best for selecting the appropriate
number of clusters?
A. The Elbow Method
B. Linear Discriminant Analysis
C. Principal Component Analysis
D. Hierarchical Clustering
Correct ANS: A. The Elbow Method
Rationale: The Elbow Method involves plotting the WCSS
against the number of clusters and selecting the k at which the
marginal decrease sharply levels off (“the elbow”), indicating an
optimal trade‑off between fit and complexity.
©2025
