Databricks Certified Machine Learning Professional Practice Test 95
Question 940
A production ML review contains this situation: A real-time feature job updates user purchase totals without checking whether upstream transaction events are delivered twice. What is the most material immediate risk in this situation?
- Scores pull the wrong customer's features or fail after name changes.
- Historical training datasets cannot be faithfully reconstructed after the overwrite.
- Several models change their effective input distributions without deliberate retraining or review.
- One purchase increases the spending feature twice, affecting downstream risk scores.
Explanation
Correct Answer
D:This consequence follows from the specific unresolved condition. One purchase increases the spending feature twice, affecting downstream risk scores. The causal constraint is: Streaming derived features should deduplicate at a stable business-event key before non-idempotent aggregation.
Incorrect Answers
A:That direct consequence is directed at a different advanced feature store concepts situation: A low-latency feature table uses external customer names as lookup keys; names change and are not unique. The more immediate risk here is one purchase increases the spending feature twice, affecting downstream risk scores.
B:That direct consequence is directed at a different advanced feature store concepts situation: A model's feature generation step overwrites all feature-table history daily even though time-aware training requires earlier snapshots. The more immediate risk here is one purchase increases the spending feature twice, affecting downstream risk scores.
C:That direct consequence is directed at a different advanced feature store concepts situation: A team updates a shared on-demand feature function without version control and unexpectedly changes both batch retraining and online serving behavior. The more immediate risk here is one purchase increases the spending feature twice, affecting downstream risk scores.
Question 941
The engineering team encounters the following issue: A churn experiment separately scripts feature scaling in training but omits that scaling when the deployed prediction job loads the classifier. Which technical principle most directly explains the root of the issue?
- Training-set fit does not establish out-of-sample predictive quality or cold-start behavior.
- VectorAssembler is a Transformer that creates a vector column, not a trainable predictor.
- Keeping fitted preprocessing and the estimator in one saved PipelineModel prevents training-serving transformation drift.
- Distributed batch inference can apply a saved SparkML PipelineModel directly to a DataFrame without collecting every row.
Explanation
Correct Answer
C:The technical constraint is specific to the reported failure. Keeping fitted preprocessing and the estimator in one saved PipelineModel prevents training-serving transformation drift. To test it, compare the logged training transformation graph with the batch scoring graph for identical raw inputs.
Incorrect Answers
A:That underlying principle is directed at a different using spark ml situation: A recommender experiment computes all evaluation metrics on training interactions and reports the result as generalization to new customers. The current incident instead turns on keeping fitted preprocessing and the estimator in one saved PipelineModel prevents training-serving transformation drift.
B:That underlying principle is directed at a different using spark ml situation: A team wants to convert several numeric inputs into one features vector but treats VectorAssembler as a clustering Estimator that should output predictions. The current incident instead turns on keeping fitted preprocessing and the estimator in one saved PipelineModel prevents training-serving transformation drift.
D:That underlying principle is directed at a different using spark ml situation: A logistics model must score 60 million daily routes already in Delta but the scoring script collects each route into Python before calling predict. The current incident instead turns on keeping fitted preprocessing and the estimator in one saved PipelineModel prevents training-serving transformation drift.
Question 942
Consider the following fictional Databricks production case: A fraud model predicts instantly but labels arrive 21 days later; an inference dashboard treats unlabelled transactions as negative examples. What consequence most specifically follows from leaving the issue unchanged?
- The affected depot stays poorly served while global metrics appear stable.
- Observability becomes a channel for customer-level data leakage.
- Material drift remains unaddressed because nobody receives the notification.
- False optimism or pessimism distorts the model's apparent accuracy and promotion decisions.
Explanation
Correct Answer
D:This consequence follows from the specific unresolved condition. False optimism or pessimism distorts the model's apparent accuracy and promotion decisions. The causal constraint is: True recall and precision require mature observed labels, not fabricated negatives for pending outcomes.
Incorrect Answers
A:That direct consequence is directed at a different drift detection and lakehouse monitoring situation: A global model predicts late deliveries accurately overall, but all mistakes occur at one small coastal depot and the team never slices its results by site. The more immediate risk here is false optimism or pessimism distorts the model's apparent accuracy and promotion decisions.
B:That direct consequence is directed at a different drift detection and lakehouse monitoring situation: A monitor owner writes a custom metric that logs unaggregated account-level health data in a broadly shared results table. The more immediate risk here is false optimism or pessimism distorts the model's apparent accuracy and promotion decisions.
C:That direct consequence is directed at a different drift detection and lakehouse monitoring situation: A fraud-feature drift monitor crosses its threshold but the alert has no assigned recipient and no incident ownership. The more immediate risk here is false optimism or pessimism distorts the model's apparent accuracy and promotion decisions.
Question 943
A production ML review contains this situation: A new fraud classifier improves overall F1 but doubles false negatives for high-value payment accounts, and an automated policy chooses it on F1 alone. Which diagnostic check provides the strongest direct evidence?
- Compare baseline seasonality, sample sizes and repeated drift statistics before each retrain event.
- Compare source completeness, freshness and drift metrics across the same incident window.
- Compare champion and candidate per-segment false-negative rates on equivalent untouched evaluation cohorts.
- Inspect prediction timestamps, label arrival dates and available labeled cohort coverage.
Explanation
Correct Answer
C:This is the most discriminating check for the observed symptoms. Compare champion and candidate per-segment false-negative rates on equivalent untouched evaluation cohorts. The underlying technical issue is that champion selection must enforce critical segment-level guardrails in addition to aggregate ranking metrics.
Incorrect Answers
A:That diagnostic evidence is directed at a different automated retraining situation: An automated retraining job triggers whenever a numerical feature mean shifts by 0.1, regardless of normal seasonal changes or sample size. For the stated incident, the decisive observation is compare champion and candidate per-segment false-negative rates on equivalent untouched evaluation cohorts.
B:That diagnostic evidence is directed at a different automated retraining situation: A feature stream loses half its source events and a retraining policy treats the resulting input shift as proof that the learned concept changed. For the stated incident, the decisive observation is compare champion and candidate per-segment false-negative rates on equivalent untouched evaluation cohorts.
D:That diagnostic evidence is directed at a different automated retraining situation: A credit model's true-label feedback arrives 45 days late, yet the retraining scheduler assumes yesterday's unlabeled predictions provide yesterday's observed accuracy. For the stated incident, the decisive observation is compare champion and candidate per-segment false-negative rates on equivalent untouched evaluation cohorts.
Question 944
Consider the following fictional Databricks production case: A hyperparameter search operates on large Spark-managed feature tables but each candidate model is small and independent, and workers repeatedly convert the entire table for each Ray task. Which result is the strongest acceptance criterion after remediation?
- A measured end-to-end trial batch reuses data without redundant full-table copies.
- Each source record receives exactly one score from its assigned region-specific model version.
- Task throughput scales with added worker capacity while the driver remains within its limits.
- Each parallel trial resolves to distinct reproducible metrics and artifact storage.
Explanation
Correct Answer
A:This verification measures the required result rather than just process liveness. A measured end-to-end trial batch reuses data without redundant full-table copies. It confirms whether the targeted action succeeded: prepare a versioned reusable feature snapshot and schedule Ray trials against appropriately shared data.
Incorrect Answers
B:That acceptance test is directed at a different scaling and tuning situation: A batch scoring pipeline trains one per-region model and then joins model predictions back using region alone, duplicating rows because each region contains multiple versioned models. Here the actual success criterion is a measured end-to-end trial batch reuses data without redundant full-table copies.
C:That acceptance test is directed at a different scaling and tuning situation: A feature-engineering job is CPU-bound on hundreds of independent daily partitions and a team upgrades only the driver to a much larger machine. Here the actual success criterion is a measured end-to-end trial batch reuses data without redundant full-table copies.
D:That acceptance test is directed at a different scaling and tuning situation: An Optuna optimization reuses the same MLflow run ID for concurrent trials, causing overlapping parameter values and artifacts. Here the actual success criterion is a measured end-to-end trial batch reuses data without redundant full-table copies.
Question 945
Review this production model incident: A customer-risk feature needs a ratio of current debt to income, but online serving recomputes it differently from the training pipeline. What end-to-end verification is most persuasive here?
- A scheduled run publishes the expected governed feature-table version with traceable input versions.
- The same representative raw inputs yield matching feature values in training and online requests.
- The model input is missing or flagged stale when no eligible feature exists inside the authorized window.
- A controlled source update appears in the online store within the documented freshness bound.
Explanation
Correct Answer
B:This verification measures the required result rather than just process liveness. The same representative raw inputs yield matching feature values in training and online requests. It confirms whether the targeted action succeeded: centralize the approved on-demand ratio transformation and its edge-case semantics.
Incorrect Answers
A:That acceptance test is directed at a different advanced feature store concepts situation: A team manually reruns a feature notebook each week before training; feature definitions change without an execution history or stable destination. Here the actual success criterion is the same representative raw inputs yield matching feature values in training and online requests.
C:That acceptance test is directed at a different advanced feature store concepts situation: A sensor classifier silently reuses features from a year-old device after months without readings, despite a maximum allowed lookback of seven days. Here the actual success criterion is the same representative raw inputs yield matching feature values in training and online requests.
D:That acceptance test is directed at a different advanced feature store concepts situation: A team publishes an online feature table but neglects its refresh schedule while the predictive model assumes values no older than 15 minutes. Here the actual success criterion is the same representative raw inputs yield matching feature values in training and online requests.
Question 946
The engineering team encounters the following issue: A bundle references a model registry name without qualifying its environment, and a development deployment accidentally updates production registered-model aliases. Which experiment most directly demonstrates a successful fix?
- A dry-run plan for dev resolves only development assets; prod changes require separate approval.
- Load tests show production SLAs hold while research workloads respect assigned capacity limits.
- Region-specific test requests reach only authorized feature storage and compliant endpoints.
- A development identity fails negative write tests against production assets while normal dev jobs succeed.
Explanation
Correct Answer
A:This verification measures the required result rather than just process liveness. A dry-run plan for dev resolves only development assets; prod changes require separate approval. It confirms whether the targeted action succeeded: parameterize catalog, schema, experiment and endpoint identifiers with validation gates per environment.
Incorrect Answers
B:That acceptance test is directed at a different environment architectures situation: A production deployment scales to 30 teams but all jobs share one compute pool with no capacity limits or cost attribution. Here the actual success criterion is a dry-run plan for dev resolves only development assets; prod changes require separate approval.
C:That acceptance test is directed at a different environment architectures situation: A multilingual prediction service must handle regional data residency, but the team copies a shared feature cache across regions without governance. Here the actual success criterion is a dry-run plan for dev resolves only development assets; prod changes require separate approval.
D:That acceptance test is directed at a different environment architectures situation: A company operates development and production models in one workspace under the same broad service account and frequently overwrites production tables during experiments. Here the actual success criterion is a dry-run plan for dev resolves only development assets; prod changes require separate approval.
Question 947
The engineering team encounters the following issue: A credit model's true-label feedback arrives 45 days late, yet the retraining scheduler assumes yesterday's unlabeled predictions provide yesterday's observed accuracy. Which technical principle most directly explains the root of the issue?
- Champion selection must enforce critical segment-level guardrails in addition to aggregate ranking metrics.
- Model candidates need evaluation on the same versioned holdout and metric definition for fair comparison.
- Production model selection must satisfy deployment constraints as well as validation quality.
- Ground-truth delay constrains performance-triggered retraining; fresh unlabeled predictions cannot establish predictive accuracy.
Explanation
Correct Answer
D:The technical constraint is specific to the reported failure. Ground-truth delay constrains performance-triggered retraining; fresh unlabeled predictions cannot establish predictive accuracy. To test it, inspect prediction timestamps, label arrival dates and available labeled cohort coverage.
Incorrect Answers
A:That underlying principle is directed at a different automated retraining situation: A new fraud classifier improves overall F1 but doubles false negatives for high-value payment accounts, and an automated policy chooses it on F1 alone. The current incident instead turns on ground-truth delay constrains performance-triggered retraining; fresh unlabeled predictions cannot establish predictive accuracy.
B:That underlying principle is directed at a different automated retraining situation: A retraining pipeline ranks candidate models using metrics calculated on different dataset versions, then selects the numerically highest result. The current incident instead turns on ground-truth delay constrains performance-triggered retraining; fresh unlabeled predictions cannot establish predictive accuracy.
C:That underlying principle is directed at a different automated retraining situation: An AutoML training job always chooses the model with the lowest training loss, even when the deployed service has a strict memory and P95 latency budget. The current incident instead turns on ground-truth delay constrains performance-triggered retraining; fresh unlabeled predictions cannot establish predictive accuracy.
Question 948
The engineering team encounters the following issue: An ML owner adds a custom metric computed from the average of a string-typed currency column without validating conversion. Which experiment most directly demonstrates a successful fix?
- Known labeled fixtures join one-to-one with their source predictions and yield the expected error rate.
- A synthetic slow drift triggers the fixed-baseline comparison before the model exceeds tolerance.
- Undersized cohorts are appropriately suppressed while adequately powered authorized slices remain measurable.
- Valid and malformed synthetic records yield explicitly documented metric results.
Explanation
Correct Answer
D:This verification measures the required result rather than just process liveness. Valid and malformed synthetic records yield explicitly documented metric results. It confirms whether the targeted action succeeded: cast and validate the intended numeric values before calculating the approved custom aggregate.
Incorrect Answers
A:That acceptance test is directed at a different drift detection and lakehouse monitoring situation: A model performance dashboard compares this week's labels to last week's predictions without joining by stable request ID or consistent outcome horizon. Here the actual success criterion is valid and malformed synthetic records yield explicitly documented metric results.
B:That acceptance test is directed at a different drift detection and lakehouse monitoring situation: A streaming payments monitor always compares today with the previous hour, and a slow week-long distribution drift escapes because adjacent hours differ only slightly. Here the actual success criterion is valid and malformed synthetic records yield explicitly documented metric results.
C:That acceptance test is directed at a different drift detection and lakehouse monitoring situation: A demographic fairness review uses a tiny sensitive subgroup with three labeled examples and reports its observed error as definitive. Here the actual success criterion is valid and malformed synthetic records yield explicitly documented metric results.
Question 949
A production ML review contains this situation: A model version is registered but the deployment job points to an old alias after promotion and continues serving last week's model. Which check shows that this particular problem has been resolved?
- Multiple clean sessions produce the same approved encoded indices for every fixture category.
- A serving query and endpoint configuration both identify the authorized promoted model version.
- The defect is caught by development unit tests before any deployment and production smoke remains clean.
- All six consumers use one pinned helper version and identical threshold fixtures.
Explanation
Correct Answer
B:This verification measures the required result rather than just process liveness. A serving query and endpoint configuration both identify the authorized promoted model version. It confirms whether the targeted action succeeded: add a release integration assertion for the expected immutable model version at the endpoint.
Incorrect Answers
A:That acceptance test is directed at a different validation testing situation: A notebook's label encoder produces a different category order depending on an unordered Python set, but tests assert only output length. Here the actual success criterion is a serving query and endpoint configuration both identify the authorized promoted model version.
C:That acceptance test is directed at a different validation testing situation: A developer changes a feature transform in a notebook and directly uses production data as the only test case. Here the actual success criterion is a serving query and endpoint configuration both identify the authorized promoted model version.
D:That acceptance test is directed at a different validation testing situation: An engineering group duplicates a validation helper across six notebooks, and two copies enforce outdated threshold values. Here the actual success criterion is a serving query and endpoint configuration both identify the authorized promoted model version.
Question 950
Consider the following fictional Databricks production case: A retailer's features arrive two hours late, and offline training uses event time alone even though the online system could not have seen those late arrivals at prediction time. What underlying principle determines the most appropriate interpretation?
- Point-in-time correctness may require both event time and actual feature availability when ingestion latency matters.
- Low-latency serving requires online feature access with a documented freshness and lookup contract.
- On-demand feature serving must account for dependency latency and failure budgets within the endpoint SLA.
- Fault-tolerant real-time feature computation requires durable streaming progress and suitable sink idempotency.
Explanation
Correct Answer
A:The technical constraint is specific to the reported failure. Point-in-time correctness may require both event time and actual feature availability when ingestion latency matters. To test it, compare feature event time, publish time and prediction decision time for a delayed fixture.
Incorrect Answers
B:That underlying principle is directed at a different advanced feature store concepts situation: A pricing endpoint queries a large offline Delta feature table for every user request, and P95 response time reaches several seconds. The current incident instead turns on point-in-time correctness may require both event time and actual feature availability when ingestion latency matters.
C:That underlying principle is directed at a different advanced feature store concepts situation: A streaming recommender's request-time feature function calls a slow third-party API synchronously on every prediction. The current incident instead turns on point-in-time correctness may require both event time and actual feature availability when ingestion latency matters.
D:That underlying principle is directed at a different advanced feature store concepts situation: An online behavioral feature counter consumes a streaming source but does not retain checkpoint state after worker restarts. The current incident instead turns on point-in-time correctness may require both event time and actual feature availability when ingestion latency matters.