GARP RAI - Risk and Artificial Intelligence Exam
Page: 2 / 30
Total 150 questions
Question #6 (Topic: Exam A)
A hospital is considering a consultant’s recommendation for a system to classify tumors as benign, malignant, or pre-malignant. The classification will be influenced by patient characteristics as well as tumor-specific information such as cell size, cell shape, etc.
Which method would be most suitable for this task?
Which method would be most suitable for this task?
A. Principal Component Analysis
B. Linear Discriminant Analysis
C. Linear Regression
D. Logistic Regression
Answer: B
Question #7 (Topic: Exam A)
An analyst is developing a model to classify daily retail transactions data into different categories. Only a small fraction of the data has been manually reviewed and labeled. The analyst wants to take advantage of the full dataset without manually labeling each transaction.
Which of the following approaches is most appropriate?
Which of the following approaches is most appropriate?
A. Unsupervised learning, because it can group transaction patterns without labels.
B. Semi-supervised learning, because it uses both labeled data and unlabeled data.
C. Reinforcement learning, because it can learn through repeated daily interaction.
D. Supervised learning, because it can predict labels from unlabeled data.
Answer: B
Question #8 (Topic: Exam A)
Word2Vec has different architectures based on how they use context and target words. Suppose you are analyzing a sentence and want to predict a missing word based on its surrounding context.
Which Word2Vec architecture would be best suited for this task?
Which Word2Vec architecture would be best suited for this task?
A. N-grams.
B. Skip-gram.
C. CBoW.
D. RNN.
Answer: C
Question #9 (Topic: Exam A)
A credit risk analyst is preparing data for a K-nearest neighbors (KNN) model used to flag potentially risky loan applications. Two features are selected:
Annual Income (dollars, X-axis)
Debt-to-Income (DTI) Ratio (percentage, Y=axis)
Before modeling, the analyst visualizes the raw data shown below.
Annual income ranges from $30,000 to $150,000, while DTI ratios range from 10% to 50%. The analyst intends to use Euclidean distance to identify similar borrowers.
Given the pattern shown in the plot, which data preparation step would be most appropriate before fitting the model?
Annual Income (dollars, X-axis)
Debt-to-Income (DTI) Ratio (percentage, Y=axis)
Before modeling, the analyst visualizes the raw data shown below.
Annual income ranges from $30,000 to $150,000, while DTI ratios range from 10% to 50%. The analyst intends to use Euclidean distance to identify similar borrowers.
Given the pattern shown in the plot, which data preparation step would be most appropriate before fitting the model?
A. Use the raw variables because KNN is robust to differences in feature scale.
B. Scale both variables so that income and DTI contribute comparably to distance calculations.
C. Apply a logarithmic transformation only to Debt-to-Income (DTI) Ratio.
D. Apply categorical encoding to both variables to eliminate scale differences.
Answer: B
Question #10 (Topic: Exam A)
An online broker is building a product recommendation system for its clients. The system initially suggests random products and refines its suggestions over time using reinforcement learning.
Which of the following statements about “ε” in reinforcement learning is accurate?
Which of the following statements about “ε” in reinforcement learning is accurate?
A. ε represents the random number between 0 and 1 that is generated during each suggestion.
B. When ε is low, the system tends to explore more often by recommending random products.
C. ε varies systematically over time, starting at 0 and gradually increasing to 1.
D. ε is a hyperparameter in reinforcement learning that controls the action of the system.
Answer: D
