Free CFA Quantitative Methods Practice Questions & Answers
1,200 exam-style Quantitative Methods questions. Pick your answer, hit Check answer, and see the worked solution — free to start, no signup.
100% free · No login to start
Question 421
A confusion matrix includes:
Select an option first.
Correct answer: C — TP, FP, FN, TN
Explanation: Why C is right: It lists all four outcomes of binary classification. Why others are wrong: A/B/D: Incomplete.
Question 422
Precision is defined as:
Select an option first.
Correct answer: B — TP / (TP + FP)
Explanation: Why B is right: Precision measures how many predicted positives are correct. Why others are wrong: A = recall C = specificity D = false discovery rate
Question 423
Recall is defined as:
Select an option first.
Correct answer: A — TP / (TP + FN)
Explanation: Why A is right: Recall measures how many actual positives are detected. Why others are wrong: B = precision C = specificity D = error rate
Question 424
The F1 score is the:
Select an option first.
Correct answer: B — Harmonic mean of precision and recall
Explanation: Why B is right: F1 balances precision and recall using the harmonic mean. Why others are wrong: A/C/D: Not the F1 definition.
Question 425
Accuracy is misleading in imbalanced datasets because:
Select an option first.
Correct answer: B — It can be high even if minority class is poorly predicted
Explanation: Why B is right: A model predicting only the majority class can still show high accuracy. Why others are wrong: A: Accuracy includes FP. C/D: Not true.
Get the full CFA question bank — free
Drop your email and we'll send you fresh CFA practice questions, fully worked solutions and exam-deadline reminders. No spam — unsubscribe in one click.
Explanation: Why A is right: Removing noisy or irrelevant tokens improves model performance. Why others are wrong: B: Usually reduces vocabulary. C/D: Not feature selection.
Question 427
Removing tokens with extremely low or high TF is:
Select an option first.
Correct answer: A — Frequency‑based feature selection
Explanation: Why A is right: TF thresholds are a classic frequency‑based pruning method. Why others are wrong: B: It reduces dimensions. C/D: Not related.
Question 428
Tokens with high mutual information values are typically:
Select an option first.
Correct answer: B — Strong predictors
Explanation: Why B is right: High MI means the token strongly helps distinguish between classes. Why others are wrong: A: Opposite of MI’s purpose. C: MI does not depend on rarity. D: Stop words usually have low MI.
Question 429
Chi‑square feature selection is used to:
Select an option first.
Correct answer: A — Test independence between token and class
Explanation: Why A is right: Chi‑square tests whether token occurrence is related to class labels. Why others are wrong: B/C/D: Not related to chi‑square.
Question 430
A major limitation of bag‑of‑words models is that they:
Select an option first.
Correct answer: A — Ignore word order
Explanation: Why A is right: Bag‑of‑words discards sequence information. Why others are wrong: B: Not required. C: TF is easily computed. D: Punctuation is handled in preprocessing.
Question 431
A key limitation of TF‑IDF is that it:
Select an option first.
Correct answer: A — Ignores word order
Explanation: Why A is right: TF‑IDF treats text as unordered counts. Why others are wrong: B: TF‑IDF is simple. C/D: It can handle both.
Question 432
ROC curves plot:
Select an option first.
Correct answer: B — True positive rate vs false positive rate
Explanation: Why B is right: ROC curves visualize classification performance across thresholds. Why others are wrong: A: That’s a PR curve. C/D: Not ROC.
Question 433
A model with high training AUC but low validation AUC is:
Select an option first.
Correct answer: B — Overfitting
Explanation: Why B is right: Large performance gap indicates memorization of training data. Why others are wrong: A: Underfitting performs poorly everywhere. C/D: Not applicable.
Question 434
Grid search is a method for:
Select an option first.
Correct answer: A — Trying combinations of hyperparameters
Explanation: Why A is right: Grid search systematically tests parameter combinations. Why others are wrong: B/C/D: Not tuning methods.
Question 435
Class imbalance most directly affects:
Select an option first.
Correct answer: A — Precision and recall
Explanation: Why A is right: Imbalance distorts error rates for minority classes. Why others are wrong: B/C/D: Not impacted.
Question 436
The primary purpose of Exploratory Data Analysis (EDA) is to:
Select an option first.
Correct answer: B — Understand patterns and distributions in the data
Explanation: Why B is right: EDA helps you understand structure, trends, and relationships before modeling. Why others are wrong: A: Training comes later. C: Cleaning is separate. D: Tuning happens after modeling.
Question 437
Feature selection aims to:
Select an option first.
Correct answer: A — Remove irrelevant or redundant features
Explanation: Why A is right: Feature selection simplifies the model and reduces noise. Why others are wrong: B: Opposite effect. C: That’s feature engineering. D: Not the goal.
Question 438
Feature engineering aims to:
Select an option first.
Correct answer: A — Create new features that improve model performance
Explanation: Why A is right: Engineering transforms or combines features to improve predictive power. Why others are wrong: B/C/D: Not feature engineering.
Question 439
A heat map is most useful for visualizing:
Select an option first.
Correct answer: A — Correlation strength
Explanation: Why A is right: Heat maps color‑code correlation values for easy interpretation. Why others are wrong: B: Not shown in heat maps. C: Missingness heat maps are separate. D: Not typical.
Question 440
A feature with extremely low variance is likely:
Select an option first.
Correct answer: A — Not useful for prediction
Explanation: Why A is right: Low‑variance features don’t help distinguish outcomes. Why others are wrong: B: Opposite. C/D: Not implied.
Question 441
A feature that is constant for all observations:
Select an option first.
Correct answer: A — Should be removed
Explanation: Why A is right: A constant feature provides no information. Why others are wrong: B: It has zero predictive value. C/D: No benefit.
Question 442
One risk of excessive feature engineering is:
Select an option first.
Correct answer: A — Overfitting
Explanation: Why A is right: Too many engineered features can cause the model to memorize noise. Why others are wrong: B: Opposite issue. C/D: Not direct effects.
Question 443
A feature strongly correlated with the target variable is:
Select an option first.
Correct answer: A — Potentially useful
Explanation: Why A is right: Strong correlation suggests predictive value. Why others are wrong: B/C/D: Not necessarily true.
Question 444
A feature that leaks information from the future is:
Select an option first.
Correct answer: A — Invalid and must be removed
Explanation: Why A is right: Using future information creates unrealistic performance and invalidates the model. Why others are wrong: B: It’s predictive only because it cheats. C/D: Not relevant.
Question 445
A feature that is highly predictive but unstable across samples may cause:
Select an option first.
Correct answer: A — Overfitting
Explanation: Why A is right: Unstable features cause the model to memorize noise. Why others are wrong: B: Underfitting = too simple. C/D: Not related.
Question 446
The main purpose of EDA is to:
Select an option first.
Correct answer: A — Understand the data before modeling
Explanation: Why A is right: EDA helps you understand structure, patterns, and issues. Why others are wrong: B: Training comes later. C/D: Not the primary goal.
Question 447
A categorical variable with only one category for all observations should be:
Select an option first.
Correct answer: A — Removed
Explanation: Why A is right: A constant feature provides no predictive information. Why others are wrong: B/C/D: No benefit.
Question 448
A feature that is difficult to interpret but improves accuracy is:
Select an option first.
Correct answer: A — Still potentially valuable
Explanation: Why A is right: Interpretability is helpful, but predictive power matters too. Why others are wrong: B/C/D: Not necessarily true.
Question 449
A feature that is highly predictive but extremely unstable across different samples is likely causing:
Select an option first.
Correct answer: A — Overfitting
Explanation: Why A is right: Instability across samples indicates the model is memorizing noise. Why others are wrong: B: Underfitting = too simple. C/D: Not relevant.
Question 450
Clustering is best described as a method that:
Select an option first.
Correct answer: B — Groups similar observations without labels
Explanation: Why B is right: Clustering is an unsupervised method that groups data based on similarity. Why others are wrong: A: That’s supervised learning. C/D: Not the definition of clustering.
Every question on this page stays free and open either way.
Add Your Testimonial
Privacy Preference
We use cookies to enhance your browsing experience, serve personalized content, and analyze our traffic. By clicking "Accept All", you consent to our use of cookies on this platform.
Before you go, can we ask you one thing?
We are just getting started, and knowing what is holding you back would help us a lot. It takes about 20 seconds.
Overall rating
★★★★★
How likely are you to recommend us to a friend? (0 to 10)What is stopping you from creating a free account?
🎉
Thank you
We read every response. This is exactly how we make PCC better for you.
CCarr
Free assistant · no signup
Carr only reads what is public on PCC, plus your own practice progress.