Free SAS Viya Supervised Machine Learning Pipelines Practice Questions

Sample Practice Questions

Try these 10 sample questions from the SAS Viya Supervised Machine Learning Pipelines practice bank. Each includes the answer and explanation. Continue with the interactive practice above to review the full question set.

1. 1A CSV file is on an analyst's laptop and has not been loaded to the server. While choosing data for a new Model Studio project, which action imports that local file?

   1. A.Select an existing in-memory table from Available.
   2. B.Use the Import tab in the Choose Data window.Correct answer: B
   3. C.Select a previously registered server table from Data Sources.
   4. D.Replace the project source with an already available CAS table.

   Explanation: The guide instructs: "The Choose Data window appears. Select the Import tab." It then says to drag the desired local data set into that window. This imports the laptop file; selecting or replacing an already available server table does not upload the local file. Data Sources likewise addresses a server-side source rather than the local-file import operation.

   Reference: SAS Viya: Machine Learning User’s Guide — Chapter 4, Importing Data, p. 96

2. 2Every training record contains input measurements and a known outcome. A model learns their mapping and predicts outcomes for new records. Which learning setting is described?

   1. A.Unsupervised learning
   2. B.Supervised learningCorrect answer: B
   3. C.Reinforcement learning
   4. D.Semisupervised learning using a mostly unlabeled training set

   Explanation: SAS states: "Supervised learning algorithms are trained using labeled examples, such as an input where the desired output is known." The described input/outcome pairs are labeled examples. Unsupervised learning lacks historical labels; reinforcement learning uses actions and rewards; the semisupervised alternative specifically relies on a mixture dominated by unlabeled records.

   Reference: SAS: Machine Learning—What It Is and Why It Matters — What are some popular machine learning methods? Supervised learning

3. 3Test data have already been used repeatedly to compare pipelines. Why does the SAS guidance recommend a further holdout set?

   1. A.To assess on fresh data after the earlier test data have influenced model selection.Correct answer: A
   2. B.To choose the next set of hyperparameters using the holdout responses.
   3. C.To add the holdout rows to the fitting sample before assessing those same rows.
   4. D.To repeat assessment on the same test rows under a different partition name.

   Explanation: SAS explains: "Since test data was used in model comparison, using it again in comparing different pipelines might introduce bias." A fresh holdout safeguards against that reuse. Tuning on it, fitting on it before assessing it, or merely renaming reused test rows would not preserve an independent final assessment.

   Reference: SAS Viya: Machine Learning Advanced Topics — Chapter 1, Holdout Data, p. 8

4. 4A new Model Studio project is created using the Blank template. What does its initial pipeline contain?

   1. A.A Data node and a Model Comparison node.
   2. B.A Data node and a Decision Tree node.
   3. C.Only the Data node.Correct answer: C
   4. D.A Data node and an Impute node.

   Explanation: The guide states: "The Blank template is the baseline pipeline that is pre-populated with only the Data node." Imputation, decision-tree modeling and model comparison can be added to a pipeline, but none is prepopulated by this template.

   Reference: SAS Viya: Machine Learning User’s Guide — Chapter 3, Creating a New Project, p. 77

5. 5An analyst fits candidate models, selects among them, and then wants one final assessment after selection. Which allocation of data roles follows the documented training/validation/test approach?

   1. A.Fit on test data, select on training data, assess on validation data.
   2. B.Fit and select on test data, then assess on training data.
   3. C.Fit on training data, select on validation data, assess finally on test data.Correct answer: C
   4. D.Fit on validation data, select on test data, assess finally on training data.

   Explanation: The User’s Guide says training data is "used for preliminary model fitting" and "The test data set is used for a final assessment of the model." Advanced Topics describes validation as being used "for the purpose of selecting variables and adjusting parameters." These roles support fitting on training, choosing among candidates on validation and assessing finally on test. The other allocations use held-out data for development and then assess using data involved in fitting.

   Reference: SAS Viya: Machine Learning User’s Guide — Chapter 2, Charitable Giving Example, Modify Variables, p. 9

6. 6For three interval-target observations, actual values are 2, 4 and 6, and predictions are 1, 4 and 8. What is the average squared error?

   1. A.1.
   2. B.√(5/3), approximately 1.291.
   3. C.5/3, approximately 1.667.Correct answer: C
   4. D.5.

   Explanation: SAS defines ASE: "The average squared error is the sum of squared errors on the final model divided by the number of observations." Here ASE = [(2−1)²+(4−4)²+(6−8)²]/3 = (1+0+4)/3 = 5/3 ≈ 1.667 target-units squared. The value 1 averages absolute errors, √(5/3) takes the root, and 5 omits division by three.

   Reference: Exploring SAS Viya: Data Mining and Machine Learning — Chapter 5, Forest Model Results, p. 105

7. 7An analyst plans several pipelines in one Model Studio project and wants to use an additional table in a node that supports it. Which design follows the project data-source model?

   1. A.Assign a different project-level data source to each pipeline.
   2. B.Assign separate project-level sources for training and validation.
   3. C.Make both tables simultaneous default project data sources.
   4. D.Use one project data source and let the supporting node reference the additional table.Correct answer: D

   Explanation: The guide specifies: "A project can have only one data source. However, some nodes can point to additional data sources." The extra table belongs to the node's supported reference, not to a second project default, a separate default for each pipeline, or a second project-level partition source.

   Reference: SAS Viya: Machine Learning User’s Guide — Chapter 3, Creating a New Project, p. 77

8. 8A highly complex model predicts its training records almost perfectly but performs poorly on genuinely new records. Which diagnosis best fits this pattern?

   1. A.OverfittingCorrect answer: A
   2. B.A successful reduction in generalization error
   3. C.An absence of model variance
   4. D.A lack of model complexity

   Explanation: The advanced guide states: "A model that is complex enough to perfectly fit the existing data might not generalize well when used to score new observations. This is referred to as overfitting." Poor new-record performance is not a successful reduction in generalization error. The pattern does not establish absent variance, and attributing it to insufficient complexity reverses the stated diagnosis.

   Reference: SAS Viya: Machine Learning Advanced Topics — Chapter 1, Overfitting, p. 7

9. 9A nominal target has levels E, N1 and N2, with E as the event. An observation with actual level N1 is predicted as N2. How is that error treated by Misclassification (MCE) versus Misclassification (Event)?

   1. A.Neither counts it.
   2. B.Both count it.
   3. C.Only Event counts it.
   4. D.Only MCE counts it.Correct answer: D

   Explanation: SAS states: "a non-event level that is classified as another non-event level does not count in the misclassification" for the Event measure. MCE counts every incorrectly predicted level. N1→N2 is wrong as a level prediction but remains non-event→non-event, so only MCE counts it.

   Reference: SAS Viya: Machine Learning Node Reference — Chapter 24, General Properties, p. 277

10. 10In Model Studio's Advisor Options, a numeric variable has 30 unique values and the interval cutoff is 20. Which measurement level does the documented cutoff rule assign?

    1. A.IntervalCorrect answer: A
    2. B.Nominal
    3. C.Ordinal
    4. D.Binary

    Explanation: The guide says: "If the number of unique levels exceeds the specified cutoff value, the variable is assigned the measurement level interval." Here 30 exceeds 20, so interval follows. Nominal, ordinal and binary are other measurement levels, but they are not the result of this cutoff rule under the stated condition.

    Reference: SAS Viya: Machine Learning User’s Guide — Chapter 3, Creating a New Project, Advisor Options, p. 78