Practice Library
All MCQs
Browse exam-wise, subject-wise, and country-wise MCQs with explanations.
Choose an option to check your answer.
A.
By using the classes of the k closest training observations
B.
By fitting a global linear equation
C.
By mining frequent itemsets
D.
By growing a decision tree
Show Answer
Correct Answer: A. By using the classes of the k closest training observations
Explanation:
KNN stores the training data and performs local voting at prediction time.
Choose an option to check your answer.
A.
The model cannot predict classes
B.
The model needs no training data
C.
The model uses too many parameters
D.
Interactions among features are not represented directly
Show Answer
Correct Answer: D. Interactions among features are not represented directly
Explanation:
When feature combinations carry information beyond individual effects, naive factorization can miss it.
Choose an option to check your answer.
A.
Set the entire posterior to zero
B.
Replace every feature with the class prior
C.
Exclude that feature's likelihood contribution
D.
Choose a random class
Show Answer
Correct Answer: C. Exclude that feature's likelihood contribution
Explanation:
Under the model, available feature likelihoods can still contribute.
Choose an option to check your answer.
A.
Use training accuracy only
B.
Assess both discrimination and calibration
C.
Check itemset support
D.
Inspect cluster silhouettes
Show Answer
Correct Answer: B. Assess both discrimination and calibration
Explanation:
Risk decisions depend on probability meaning, not just class ranking.
Choose an option to check your answer.
A.
Agreement between predicted probabilities and observed outcome frequencies
B.
Ordering features by importance
C.
Choosing the largest class
D.
Scaling attributes to zero mean
Show Answer
Correct Answer: A. Agreement between predicted probabilities and observed outcome frequencies
Explanation:
Among cases assigned probability 0.8, roughly 80 percent should be positive if calibrated.
Choose an option to check your answer.
A.
A support and confidence
B.
A centroid and radius
C.
A margin and kernel
D.
A mean and a variance
Show Answer
Correct Answer: D. A mean and a variance
Explanation:
The normal likelihood is determined by class-specific means and variances.
Choose an option to check your answer.
A.
They make every likelihood zero
B.
They eliminate class priors
C.
Their evidence may be counted multiple times
D.
They force continuous predictions
Show Answer
Correct Answer: C. Their evidence may be counted multiple times
Explanation:
The independence model treats correlated copies as separate evidence.
Choose an option to check your answer.
A.
It always finds nonlinear clusters
B.
Training and prediction are fast with relatively few parameters
C.
It needs no labeled data
D.
It searches every feature subset
Show Answer
Correct Answer: B. Training and prediction are fast with relatively few parameters
Explanation:
Conditional independence permits separate feature statistics per class.
Choose an option to check your answer.
A.
Accurate class ranking may still result despite imperfect probability estimates
B.
The algorithm tests dependence explicitly
C.
It removes correlated features automatically
D.
Independence is never used
Show Answer
Correct Answer: A. Accurate class ranking may still result despite imperfect probability estimates
Explanation:
Classification only requires the correct posterior ordering, not perfect probability calibration.
Choose an option to check your answer.
A.
All priors become equal automatically
B.
Minority classes receive prior one
C.
Priors are ignored during prediction
D.
Majority classes receive larger prior probabilities unless adjusted
Show Answer
Correct Answer: D. Majority classes receive larger prior probabilities unless adjusted
Explanation:
Empirical priors reflect observed class frequencies.
Choose an option to check your answer.
A.
They guarantee independent features
B.
They remove class imbalance
C.
They prevent numerical underflow and turn products into sums
D.
They convert continuous values to categories
Show Answer
Correct Answer: C. They prevent numerical underflow and turn products into sums
Explanation:
Multiplying many small probabilities can underflow to zero.
Choose an option to check your answer.
A.
Removes rare categories
B.
Adds a small count to category frequencies before probability estimation
C.
Standardizes continuous features
D.
Increases the number of classes
Show Answer
Correct Answer: B. Adds a small count to category frequencies before probability estimation
Explanation:
Add-one smoothing prevents zero conditional probabilities.
