CompTIA DY0-001: CompTIA DataAI (formerly DataX)

Last checked October 22, 2026 against the official CompTIA exam page.

1. Question 1Machine learning

   A fully grown decision tree achieves near-perfect training accuracy but much lower accuracy on held-out data. Which action most directly reduces this overfitting?

   1. AReplace the classifier with a linear regression on the labels
   2. BIncrease the maximum depth to capture more splits
   3. CLimit the maximum depth or apply cost-complexity pruning
   4. DRemove the validation set from the pipeline
   Show answer and explanationHide answer

   Correct answer: C

   Limiting depth or pruning reduces the variance of a tree by restricting how closely it fits the training data. Increasing depth makes overfitting worse, and linear regression on class labels is not a fix for variance.

2. Question 2Operations and processes

   A retailer wants its analytics warehouse to reflect inserts, updates, and deletes from its transactional database continuously, without repeated full extracts. Which ingestion approach fits this requirement?

   1. AChange data capture from the database transaction log
   2. BNightly full table dumps copied to a file share via a cron trigger
   3. CScraping the application user interface through screen text
   4. DManual CSV exports by business users on a weekly schedule
   Show answer and explanationHide answer

   Correct answer: A

   Change data capture reads committed changes from the transaction log and streams them downstream, so only deltas move. Full dumps and manual exports are batch processes that repeat full loads, and UI scraping is fragile and unreliable.

3. Question 3Mathematics and statistics

   A data scientist runs a two-sample t-test comparing conversion rates between two landing pages and obtains a p-value of 0.03 at a significance level of 0.05. Which interpretation is correct?

   1. AData this extreme would occur about 3% of the time if the two population means were equal
   2. BThere is a 97% probability that the alternative hypothesis is true given the observed data from this test
   3. CThe effect size is 3% of the baseline conversion rate, meaning the new page lifts conversions by that amount
   4. DThe test has a 3% chance of missing a real difference in this small sample size
   Show answer and explanationHide answer

   Correct answer: A

   A p-value is the probability of observing a result at least as extreme as the sample result assuming the null hypothesis is true. It does not give the probability that the null or alternative hypothesis is true, and it does not measure effect size.

4. Question 4Modeling, analysis, and outcomes

   A model reports 99% validation accuracy but performs poorly in production. Which two findings most strongly suggest data leakage? (Choose two.)

   1. AA feature is computed using information recorded after the prediction time
   2. BImputation and scaling were fitted on the full dataset before the train-validation split
   3. CTraining accuracy is slightly higher than validation accuracy
   4. DThe learning rate was reduced during the final epoch
   Show answer and explanationHide answer

   Correct answer: A, B

   Features built from post-outcome data and preprocessing fitted before splitting both let information from the validation set or future leak into training. Slightly higher training accuracy and a lower learning rate are normal and do not indicate leakage.

5. Question 5Specialized applications of data science

   A card-fraud model flags 0.2% of transactions, and only a small fraction of flagged transactions are truly fraudulent. Which evaluation metric is most informative for this class-imbalanced problem?

   1. AR-squared on the raw transaction amount
   2. BOverall accuracy on the full test set
   3. CPrecision-recall curve and average precision
   4. DMean squared error of predicted transaction amounts
   Show answer and explanationHide answer

   Correct answer: C

   With heavy class imbalance, accuracy can be high even for a useless model, while precision-recall analysis shows how many flagged cases are fraud and how many fraud cases are caught. Mean squared error and R-squared are regression measures.