MLA-C01 Practice Question 33: Fuzzy Deduplication FindMatches
AWS Certified Machine Learning Engineer – Associate (MLA-C01) · Test 4 · question 3 of 10
A retailer is building a customer lifetime value model from customer records merged out of three acquired companies. Many customers appear more than once, but the copies never match exactly: one system stores "Jon Smith, 12 Oak St.", another "Jonathan Smith, 12 Oak Street", and the three systems share no common customer ID. Duplicate customers are inflating the training set.
Which AWS capability is designed to identify these duplicate records?
Show answer and explanation
Answer:
FindMatches is an AWS Glue machine learning transform built to find duplicate or matching records even when they share no unique identifier and no fields match exactly, which is the situation here. You teach it by labeling sample pairs as match or non-match, and it learns how much variation in names and addresses still counts as the same customer (option A).
- B is wrong because DropDuplicates removes only rows whose selected columns are *identical*. "Jon" and "Jonathan", or "St." and "Street", are different strings, so these duplicates survive. - C is wrong for the same reason: SELECT DISTINCT collapses exact duplicates only. - D is wrong because k-means groups records by distance in numeric feature space. A cluster holds many *similar* but distinct customers, so keeping one record per cluster would delete real customers, and the number of clusters would have to match the unknown number of real customers.
Reference: Record matching with AWS Lake Formation FindMatches