MCQs

Disclaimer:

Educational Use Only:

The website links provided are to be used strictly for educational and knowledge-gaining purposes only.

No Misuse:

Do not misuse the links or the information contained therein for any commercial, illegal, or unethical activities.

Respect for Websites and Businesses:

Respect the intellectual property and business interests of the websites and their owners. Do not attempt to alter, copy, or redistribute content without proper authorization.

Acknowledge Value of Knowledge:

Recognize and value the knowledge and resources shared by these websites. Use the information responsibly and ethically.

No Endorsement:

The provision of these links does not imply endorsement of the linked sites, their content, or their products and services. Users are responsible for evaluating the suitability and accuracy of the information.

User Responsibility:

By accessing these links, you agree to use the information at your own risk and to comply with all applicable laws and website terms of use.

Overall MCQs for Reference

1. Which of the following best describes record linkage in the context of machine learning?
   a. Merging unrelated datasets
   b. Identifying records referring to the same entity across different sources
   c. Deleting duplicate records
   d. Assigning random IDs to data points
   Answer: bwikipedia​
2. Why are restrictions or constraints used in record linkage?
   a. To increase computational costs
   b. To reduce privacy risks and computational complexity
   c. To eliminate valid matches
   d. To maximize false positives
   Answer: bdamavis+1​
3. What is the main privacy concern in record linkage?
   a. Improved ML accuracy
   b. Unauthorized disclosure of personal data
   c. Increased database size
   d. Reduced training time
   Answer: barxiv+2​
4. Which approach combines human clerical review with automated matching to minimize information disclosure?
   a. Single-pass blocking
   b. Multi-layer active learning
   c. Brute-force search
   d. Exact match only
   Answer: barxiv​
5. What technique is commonly used to reduce the number of record comparisons for efficiency?
   a. Data augmentation
   b. Blocking
   c. Hashing
   d. Sampling
   Answer: bpublications.isi+1​
6. In privacy-preserving record linkage, what principle ensures data owners share only what is needed?
   a. Least privilege
   b. Need-to-know
   c. Open access
   d. All-or-nothing sharing
   Answer: barxiv​
7. Why can using too many blocking rules reduce matching accuracy?
   a. It increases training time
   b. It loses potential linkage candidates
   c. It reduces false negatives only
   d. It eliminates true duplicates
   Answer: btextbook.coleridgeinitiative+1​
8. Which record linkage method does NOT use optimal thresholds for matching?
   a. Probabilistic
   b. Machine learning
   c. Rule-based
   d. Hybrid
   Answer: ctextbook.coleridgeinitiative​
9. What is the main trade-off when setting match/non-match thresholds in record linkage?
   a. Speed vs. quality
   b. Recall vs. precision
   c. Data size vs. model size
   d. Accuracy vs. cost
   Answer: bwikipedia​
10. What is a significant challenge when matching large datasets during record linkage?
    a. Lack of fields
    b. Combinatorial explosion
    c. Label drift
    d. Increasing field lengths
    Answer: bdamavis+1​
11. What does “blocking” aim to preserve in record linkage?
    a. All possible links
    b. A manageable candidate set
    c. Unique identifiers
    d. Manual review only
    Answer: bpublications.isi​
12. Which of the following is a restriction for machine learning in record linkage when labels are unavailable?
    a. Only supervised learning
    b. Only unsupervised learning
    c. Both supervised and unsupervised
    d. No learning is possible
    Answer: bdamavis​
13. In probabilistic record linkage, what are compared to determine linkages?
    a. Rule sets
    b. Probability scores
    c. Exclusively unique identifiers
    d. Data storage costs
    Answer: bwikipedia+1​
14. What is a potential drawback of rule-based record linkage?
    a. Inflexibility to data changes
    b. Overfitting large datasets
    c. High training cost
    d. Needs large memory
    Answer: atextbook.coleridgeinitiative+1​
15. Why is data minimization crucial in privacy-preserving record linkage?
    a. Reduces database size
    b. Minimizes breach impact
    c. Makes models faster
    d. Increases accuracy
    Answer: bfigshare.mq+1​
16. How is “blocking” typically implemented in practice?
    a. Sorting only on one field
    b. Restricting to records with common discriminating attribute(s)
    c. Merging all records
    d. Deleting sparse entries
    Answer: bwikipedia​
17. What is “active learning” used for in privacy-preserving record linkage?
    a. Random labeling
    b. Selectively querying ambiguous candidates
    c. Maximizing false negatives
    d. Removing all human intervention
    Answer: bfigshare.mq+1​
18. Which method in record linkage is best suited for non-linear and complex data dependencies?
    a. Logistic regression
    b. Decision tree
    c. Neural networks
    d. Jaccard similarity
    Answer: clib.purdue+1​
19. What is a direct effect of overly restrictive blocking conditions?
    a. More candidate pairs
    b. Loss of true matches
    c. Increased recall
    d. Faster data entry
    Answer: bpublications.isi​
20. Which metric improves when blocking is applied correctly without missing links?
    a. Recall
    b. Positive predictive value (precision)
    c. Specificity
    d. False discovery rate
    Answer: bpublications.isi​
21. What is the main aim of the Fellegi-Sunter model in record linkage?
    a. Maximize false matches
    b. Minimize linkage errors using probabilistic modeling
    c. Increase blocking efficiency
    d. Rely solely on machine learning
    Answer: bwikipedia​
22. What does the “privacy budget” manage in differential privacy approaches to record linkage?
    a. The total data size
    b. The allowable leakage of private information
    c. The computational resources
    d. The recall rate
    Answer: bfigshare.mq​
23. Which statement is true for privacy-preserving record linkage (PPRL)?
    a. It shares all personal identifiers
    b. It restricts shared information to what is necessary for linkage
    c. It eliminates privacy concerns completely
    d. It always uses public datasets
    Answer: barxiv+1​
24. Why can creating labeled datasets for record linkage in sensitive domains be challenging?
    a. Lack of labeling tools
    b. Privacy and confidentiality constraints
    c. Absence of unique identifiers
    d. Data redundancy
    Answer: bfigshare.mq​
25. Which restriction is important when multiple records have high linkage probabilities for a given record?
    a. Block all
    b. Link all above a threshold or select best match only
    c. Reject all matches
    d. Merge all datasets
    Answer: bdamavis​
26. Which of the following is a hybrid record linkage model?
    a. Only induction
    b. Only clustering
    c. Combination of induction and clustering
    d. Only rule-based
    Answer: clib.purdue​
27. What challenge can arise if machine learning models for record linkage are trained only on synthetic data?
    a. High overfitting
    b. Bias or reduced accuracy on real data
    c. Unrestricted matches
    d. Less need for privacy
    Answer: bfigshare.mq​
28. What is the main purpose of cardinality constraints in privacy-preserving linkage?
    a. Ensure only unique matches
    b. Count distinct individuals without identity compromise
    c. Minimize the amount of training data
    d. Increase linkage errors
    Answer: bfigshare.mq​
29. Which field is most often “locked” to limit combinations during record linkage?
    a. Email address
    b. Unique, reliable identifier like CNAE code
    c. Last name
    d. Date created
    Answer: bdamavis​
30. What is a major risk when blocking is too permissive?
    a. Too many false negatives
    b. Computational overload
    c. Discarding actual matches
    d. Losing all data
    Answer: bwikipedia​
31. In privacy-preserving record linkage, which method can limit the number of entities revealed per query?
    a. k-anonymity
    b. Differential privacy
    c. Data duplication
    d. Relational algebra
    Answer: bsciencedirect+1​
32. What is “need-to-know” in the context of PPRL?
    a. Only known fields are queried
    b. Data owners determine minimal data sharing
    c. Publicly available information is maximized
    d. All records are fully disclosed
    Answer: barxiv​
33. What is one outcome of combining record linkage with unsupervised clustering in PPRL?
    a. Increased identity disclosure
    b. High linkage quality without identity compromise
    c. Always uniform cluster sizes
    d. Removal of privacy constraints
    Answer: bfigshare.mq​
34. What is the term for restricting matching to pairs that agree on specific attributes?
    a. Blocking
    b. Sampling
    c. Feature engineering
    d. Oversampling
    Answer: awikipedia​
35. What’s a potential privacy risk if blocking attributes themselves are sensitive?
    a. Increased computation
    b. Attribute leakage
    c. Higher recall
    d. Reduced cost
    Answer: barxiv+1​
36. Why is transparency important in the record linkage process?
    a. Increases linkage error
    b. Helps evaluate and understand privacy risks
    c. Reduces accountability
    d. Increases dataset sizes
    Answer: btextbook.coleridgeinitiative​
37. What disadvantage do classical record linkage methods face as dataset complexity increases?
    a. Better scalability
    b. Lack of flexibility and adaptability
    c. Greater interpretability
    d. Native data privacy support
    Answer: bdamavis​
38. Which algorithmic constraint is sometimes applied during blocking to ensure scalability?
    a. Limiting block size
    b. Requiring unique fields
    c. Denoising data
    d. Enforcing k-fold validation
    Answer: apublications.isi​
39. What type of attribute is ideal for blocking?
    a. Infrequent attribute
    b. Highly discriminating non-sensitive attribute
    c. Variable attribute
    d. Null attribute
    Answer: bwikipedia​
40. What is an advantage of multi-layer review protocols in PPRL?
    a. Reduces linkage quality
    b. Minimizes information disclosed at each review
    c. Increases cost only
    d. Simplifies implementation
    Answer: barxiv​
41. What is a benefit of using machine learning models that do not assume conditional independence among attributes during linkage?
    a. Higher accuracy when training data is sufficient
    b. Lower computational efficiency
    c. Always lower recall
    d. No improvement observed
    Answer: awikipedia​
42. What is a typical optimization goal for blocking schemes?
    a. Maximize number of candidates
    b. Maximize precision and recall trade-off
    c. Minimize linkage quality
    d. Always prefer higher recall
    Answer: bpublications.isi​
43. Why might a practitioner choose a rule-based linkage method despite its restrictions?
    a. Simpler and faster for small datasets, easily explained
    b. Best scalability
    c. Most accurate for all data types
    d. Most privacy-friendly
    Answer: atextbook.coleridgeinitiative​
44. What characteristic of machine learning makes it attractive for complex record linkage?
    a. Ability to model non-linear dependencies
    b. Simpler implementation than rule-based
    c. Lower data requirements
    d. Invariant scaling
    Answer: alib.purdue+1​
45. When is manual clerical review most needed in PPRL?
    a. For non-ambiguous matches
    b. For uncertain/ambiguous match candidates
    c. For the largest datasets
    d. For uniquely identified entities
    Answer: barxiv​
46. In data integration, what is the danger of ignoring restrictions in record linkage?
    a. Higher false negatives only
    b. Increased privacy risk and computation
    c. Higher recall
    d. Better oversight
    Answer: bsciencedirect+2​
47. Which constraint helps enforce fairness in PPRL?
    a. Random matching
    b. Fairness-constrained differential privacy
    c. Greedy clustering
    d. Removing all constraints
    Answer: bfigshare.mq​
48. What is one approach to reduce error rate in privacy-preserving unsupervised clustering for record linkage?
    a. Increase privacy budget
    b. Reduce record diversity
    c. Use novel clustering algorithms designed for PPRL
    d. Only utilize synthetic data
    Answer: cfigshare.mq​
49. What must be balanced when deciding on the information to release for linkage tasks in sensitive environments?
    a. Recall vs. speed
    b. Linkage quality and privacy risks
    c. Database size and speed
    d. Label quantity and diversity
    Answer: barxiv+1​
50. When restricting the set of comparable attributes, what outcome should be regularly assessed?
    a. The resulting precision and recall balance
    b. Always higher recall
    c. Only data diversity
    d. Increasing computation
    Answer: apublications.isi+1​

1. https://en.wikipedia.org/wiki/Record_linkage
2. https://blog.damavis.com/en/record-linkage-with-machine-learning/
3. https://arxiv.org/abs/2412.04178
4. https://figshare.mq.edu.au/articles/thesis/Privacy-Preserving_Data_Sharing_with_Machine_Learning/25039997
5. https://www.sciencedirect.com/science/article/abs/pii/S0169023X2300040X
6. https://publications.isi.edu/downloads/integration/michelson06-aaai.pdf
7. https://textbook.coleridgeinitiative.org/chap-link.html
8. https://docs.lib.purdue.edu/cgi/viewcontent.cgi?article=2572&context=cstech
9. https://www.sanfoundry.com/1000-machine-learning-questions-answers/
10. https://www.scribd.com/document/385447962/Data-Mining-and-Warehouse-MCQS-with-Answer-Good-docx
11. https://cran.r-project.org/web/packages/reclin2/vignettes/record_linkage_using_machine_learning.html
12. https://iotmumbai.bharatividyapeeth.edu/media/pdf/ETI_22618_UT1_Question_Bank_2022-23_080323.pdf
13. https://testbook.com/objective-questions/mcq-on-information-retrieval–64f9c0d1513684ddf3ab23c0
14. https://www.igmguru.com/blog/machine-learning-mcqs
15. https://ieeexplore.ieee.org/document/10356781/
16. https://www.gyansetu.in/blog/machine-learning-mcq-interview-preparation/
17. https://www.ncbi.nlm.nih.gov/books/NBK253312/
18. https://www.boozallen.com/insights/ai-research/privacy-preserving-record-linkage.html
19. https://www.niss.org/sites/default/files/winkler.pdf
20. https://dl.acm.org/doi/10.1109/TKDE.2023.3342757