Domain 3: Supervised Learning with SAS

SAS AI and Machine Learning · 61 questions

• A data scientist runs PROC HPLOGISTIC on a binary outcome in SAS. Which statement correctly sets the reference level for a CLASS variable named 'Region' to 'West'?
• A data scientist trains a random forest model using PROC HPFOREST in SAS and wants to limit the maximum number of leaf nodes in each tree to control overfitting. Which option accomplishes this?
• In PROC HPNEURAL, which activation function is specified for hidden layer neurons using the HIDDEN statement?
• A data scientist uses PROC HPREG in SAS to perform LASSO variable selection. Which option enables LASSO regularization?
• When using gradient boosting in SAS Viya's gbtreetrain CAS action, what does the SUBSAMPLERATE= parameter control?
• A decision tree trained with PROC HPSPLIT in SAS has grown too deep and is overfitting. Which pruning approach is available in PROC HPSPLIT?
• In PROC HPLOGISTIC, which fit statistic should a data scientist examine to evaluate whether adding a predictor significantly improves the model compared to a nested simpler model?
• A data scientist runs PROC HPFOREST with SEED=12345 and obtains variable importance scores. The top variable is 'creditscore'. Which metric does PROC HPFOREST use by default to compute variable importance?
• Which SAS procedure is purpose-built for training gradient boosting decision trees (GBDT) in the SAS 9.4 environment (not Viya)?
• In a SAS Viya pipeline, PROC HPLOGISTIC is used to build a binary classification model. The analyst wants to obtain predicted probabilities for the positive class. Which output statement is correct?
• In PROC HPLOGISTIC output, the 'Convergence Status' indicates 'Convergence criterion (GCONV=1E-8) satisfied.' What does this mean?
• A data scientist applies PROC HPFOREST to a regression problem. Which option in PROC HPFOREST specifies the number of variables randomly considered at each split?
• In SAS, a data scientist wants to use cost-sensitive learning to handle a class imbalance problem in PROC HPLOGISTIC. Which approach correctly applies differential misclassification costs?
• A data scientist trains a neural network using PROC HPNEURAL with 2 hidden layers. Which option in the PROC statement controls the maximum number of training iterations?
• When using PROC HPREG with forward stepwise selection, which criterion controls which variables are added to the model at each step?
• A SAS data scientist uses the following code: PROC HPFOREST DATA=train SEED=999; TARGET churn / LEVEL=NOMINAL; INPUT age income tenure / LEVEL=INTERVAL; INPUT segment / LEVEL=NOMINAL; OOB; RUN; What does the OOB statement request?
• In PROC HPSPLIT for classification trees, which impurity measure is the default splitting criterion?
• A data scientist uses PROC HPREG to fit a polynomial regression with terms up to degree 3. How should the cubic term for variable 'X' be specified in the MODEL statement?
• A SAS data scientist submits the following PROC HPNEURAL code and gets a 'CONVERGENCE NOT ACHIEVED' warning. Which adjustment is most likely to resolve convergence issues?
• A data scientist uses PROC HPLOGISTIC and requests the ROC curve output. Which ODS output object contains the ROC curve data points (sensitivity and 1-specificity pairs)?
• A data scientist in SAS Viya runs PROC HPREG and gets a Type I SS (Sequential Sum of Squares) and Type II SS (Partial Sum of Squares) table. When do the Type I and Type II SS values differ for a predictor?
• When using the SAS Viya decisionTree.dtreeTrain CAS action, which parameter controls the minimum number of training observations required before a node will be split?
• In SAS, which procedure is most appropriate for building a multivariate regression model when the number of predictors (p=500) greatly exceeds the number of observations (n=200)?
• A data scientist uses PROC HPNEURAL in SAS with TRAIN TECH=LEVMAR. What optimization algorithm does LEVMAR specify?
• In PROC HPLOGISTIC, a data scientist specifies LINK=PROBIT in the MODEL statement instead of the default LINK=LOGIT. What change does this make to the model?
• A data scientist trains a PROC HPFOREST model on a credit scoring dataset. The training AUC is 0.95 but validation AUC is 0.71. Which PROC HPFOREST option is most likely to close this gap without changing the dataset?
• In PROC HPLOGISTIC, which option generates a score dataset with predicted probabilities for each new observation?
• A data scientist uses the gradient boosting gbtreetrain CAS action in SAS Viya and wants to tune the learning rate (shrinkage) parameter. What is the effect of decreasing the learning rate from 0.1 to 0.01 while keeping all other parameters constant?
• A data scientist specifies PROC HPSPLIT with a GROW ENTROPY option. What splitting criterion does this select?
• A data scientist runs PROC HPNEURAL for regression. Which option correctly specifies that the output layer should use a linear (identity) activation function for a continuous target?
• In a SAS Viya gradient boosting model (gbtreetrain CAS action), a data scientist sets SUBSAMPLERATE=0.8 and COLSAMPLEBYTREE=0.7. What do these parameters together create?
• A data scientist trains a decision tree in PROC HPSPLIT. The output shows a leaf node with 3 observations, all of the same class. What is the Gini impurity of this node?
• A SAS data scientist is preparing features for a neural network in PROC HPNEURAL. Input variables have widely different scales: 'age' ranges 18-90, 'income' ranges 20000-500000, and 'loanamount' ranges 1000-50000. Why is standardization critical before training?
• A data scientist uses PROC HPREG with SELECTION=STEPWISE and SLENTRY=0.05 SLSTAY=0.10. What do these two significance level options control?
• In SAS Viya's gradient boosting implementation, a data scientist wants to prevent individual trees from memorizing specific training observations. Which combination of parameters directly implements this goal?
• For a binary classification problem, which splitting criterion does PROC HPFOREST use by default?
• In PROC HPREG diagnostics, a predictor has VIF = 18. What does this indicate and what is the typical remediation?
• In PROC LOGISTIC with CLASS variables, which parameterization uses the last alphabetically sorted category as the reference by default?
• A gradient boosting model with 1000 trees and learning rate 0.01 shows overfitting. Which regularization approach is most effective without reducing tree count?
• In PROC HPNEURAL, which statement and option specify the number of neurons in a hidden layer?
• Why is PROC SURVEYLOGISTIC preferred over PROC LOGISTIC for survey data analysis?
• How does PROC HPFOREST handle missing values in predictor variables during training and scoring?
• PROC HPFOREST is trained with DEPTH=10 on a dataset with 100 predictor variables. A colleague suggests reducing DEPTH=4. What is the primary effect of reducing tree depth in a random forest?
• PROC HPREG is run with SELECTION=STEPWISE and the default entry/removal significance levels. A variable 'creditscore' is added at step 2 but then removed at step 6 when 'paymenthistory' enters. At step 9, 'creditscore' re-enters after 'debtratio' removes 'paymenthistory'. What is the most appropriate modeling response to this stepwise cycling?
• PROC HPFOREST generates an OOB error estimate as part of its default output. What distinguishes the OOB error estimate from a standard holdout validation error?
• PROC GAMPL is used to model a non-linear relationship between patient age and hospital readmission. Which aspect of PROC GAMPL's penalized spline distinguishes it from including polynomial terms in PROC LOGISTIC?
• PROC LIFETEST is run with the STRATA= statement to compare survival curves between two treatment arms. By default, which test does PROC LIFETEST report?
• In PROC PHREG, a hazard ratio of 2.5 is reported for a binary predictor 'treatment' (0=control, 1=treated). How is this interpreted?
• A PROC HPNEURAL model is trained with MLPARCH(20 10) and the VALIDATION= option. Training loss decreases monotonically but validation loss begins rising at epoch 15. What does PROC HPNEURAL do when the VALIDATION= option is specified?
• In PROC GLMSELECT with SELECTION=LASSO, the PATH= option is specified. What does the solution path represent?
• A data scientist builds a PROC HPFOREST model and a PROC TREEBOOST model on the same data and compares their variable importance outputs. A colleague claims both should rank variables identically. Why is this assumption incorrect?
• A gradient boosting model is trained with SHRINKAGE=0.01 (low learning rate) and NTREES=500. A colleague trains the same model with SHRINKAGE=0.1 and NTREES=50. Assuming the same training dataset, what is the expected difference in model behavior?
• PROC HPREG is run on a dataset with 50 predictors. VIF values for three predictors are 22, 18, and 31. Which action is most appropriate based on these VIF values?
• PROC HPLOGISTIC is run for a binary fraud classification. The output shows concordance = 0.83, discordance = 0.14, and ties = 0.03. How is the c-statistic (AUC) computed from these values?
• A SAS analyst fits PROC GENMOD to count data representing hospital readmissions per patient. The output shows significant overdispersion (deviance/DF = 3.2). Which distribution family should replace Poisson to better handle overdispersion?
• PROC GLMSELECT with SELECTION=STEPWISE is run. A variable is added at step 3, removed at step 5, and added again at step 8. What does this cycling indicate?
• A SAS analyst uses PROC SURVEYSELECT to create a stratified random sample. Why is stratified sampling preferred over simple random sampling when the population contains small but important subgroups?
• PROC DISCRIM is used to classify observations into predefined groups. Which assumption does Linear Discriminant Analysis (LDA) in PROC DISCRIM make that Quadratic Discriminant Analysis (QDA) does not?
• A PROC HPFOREST model is trained with VARSTOTRY=5 on a dataset with 40 predictor variables. What does VARSTOTRY=5 control?
• A SAS developer builds a gradient boosting model using PROC TREEBOOST with the NTREES=200, SHRINKAGE=0.05, and SUBSAMPLE=0.8 options. What does SUBSAMPLE=0.8 control in stochastic gradient boosting?
• PROC LIFEREG is used to fit a Weibull accelerated failure time (AFT) model. How does PROC LIFEREG differ from PROC PHREG in its modeling approach?