PEGACPDS26V1 sample questions

Question 1

An organization is deploying the out-of-the-box Predict Web Propensity model in Pega Customer Decision Hub. The marketing team notices that a small percentage of web visitors are not receiving the AI-driven next best actions, but instead receive a random selection of actions. What is the primary purpose of this behavior?

Accepted Answer

To establish a control group that measures the true business lift generated by the AI model compared to random selection.

Answer

To reduce the computational load on the Adaptive Decision Manager during peak web traffic periods.

Answer

To test new adaptive models in shadow mode before promoting them to the champion position.

Answer

To ensure that ethical bias checks are passed by exposing all demographics to an equal distribution of offers.

Question 2

A retail bank uses Adaptive Decision Manager (ADM) to personalize credit card offers. A data scientist introduces 50 new predictors, including several highly correlated demographic fields (e.g., 'Age' and 'Years in Workforce'). How does ADM natively handle these highly correlated predictors during model learning?

Accepted Answer

ADM automatically groups highly correlated predictors together and only utilizes the best-performing predictor from that group for scoring.

Answer

ADM requires the data scientist to manually remove one of the correlated predictors via the Prediction Studio interface to prevent overfitting.

Answer

ADM applies a Principal Component Analysis (PCA) transformation to merge the correlated predictors into a single composite numerical variable.

Answer

ADM penalizes the correlated predictors by artificially lowering their Area Under the Curve (AUC) scores until they drop below the activation threshold.

Question 3

True or False: In Pega Infinity '26, adaptive models must be taken offline periodically so that a data scientist can manually retrain them using the latest batch of customer response data.

Accepted Answer

False

Answer

True

Question 4

While reviewing the Adaptive Model monitor tab in Prediction Studio, a data scientist examines the Bubble chart plotting Model Performance (AUC) on the Y-axis and Success Rate on the X-axis. Several models appear clustered in the bottom-right quadrant. Which TWO conclusions can be drawn about these specific models? (Select TWO)

```mermaid
quadrantChart
title Model Performance vs Success Rate
x-axis Low Success Rate --> High Success Rate
y-axis Low AUC --> High AUC
quadrant-1 Stars
quadrant-2 Niche
quadrant-3 Dead
quadrant-4 Cash Cows
Model A: [0.8, 0.2]
Model B: [0.9, 0.3]
```

Accepted Answer

They are effectively driving business value because they have a high rate of positive customer responses.

Accepted Answer

They are struggling to learn and may lack sufficient relevant predictors to differentiate customer behavior.

Answer

They have exceptional predictive power and are accurately distinguishing between likely responders and non-responders.

Answer

They have not yet received enough responses to activate any predictors and are still in the initial learning phase.

Answer

They are prime candidates for immediate deprecation because they generate no revenue for the business.

Question 5

When configuring an adaptive model in Prediction Studio, a data scientist must define the outcome mapping. If the business wants the model to predict the likelihood of a customer clicking a web banner, how should the outcomes be categorized?

Accepted Answer

'Click' should be defined as the target behavior, and 'Impression' (or 'Ignore') should be defined as the alternative behavior.

Answer

'Click' should be defined as the primary propensity, and 'Conversion' should be defined as the secondary propensity.

Answer

'Impression' should be defined as the target behavior, and 'Click' should be defined as the alternative behavior.

Answer

Both 'Click' and 'Impression' should be mapped as target behaviors to maximize the calculated success rate.

Question 6

A data scientist adds a new categorical predictor named 'Customer_Occupation' to an adaptive model. The organization has over 1,500 distinct occupation titles in its database. How does the Adaptive Decision Manager (ADM) process this high-cardinality predictor to ensure the model remains performant and avoids overfitting?

Accepted Answer

ADM groups the most statistically significant occupation titles into discrete bins and groups the remaining less frequent titles into a single 'Residual' or 'Other' bin.

Answer

ADM automatically rejects categorical predictors with more than 100 unique values, requiring the data scientist to manually group them prior to ingestion.

Answer

ADM assigns a unique weight to every single occupation title, resulting in a model with 1,500 individual coefficients for that predictor.

Answer

ADM converts the categorical values into numerical values using one-hot encoding, expanding the model to include 1,500 new binary predictors.

Free PEGACPDS26V1 sample questions

True or False: In Pega Infinity '26, adaptive models must be taken offline periodically so that a data scientist can manually retrain them using the latest batch of customer response data.

When configuring an adaptive model in Prediction Studio, a data scientist must define the outcome mapping. If the business wants the model to predict the likelihood of a customer clicking a web banner, how should the outcomes be categorized?

6 more free samples are waiting