PEGACPDS88V1 Certified Pega Data Scientist Practice Questions
Prepare for PEGACPDS88V1 with more than an answer.
- Exam fee
- $190 USD
- Level
- Professional
- Valid for
- Lifetime (no expiration)
Domains covered on the exam 7
- AI for Customer Decision Hub6%
- Adaptive Analytics28%
- Predictive Analytics25%
- Prediction Patterns22%
- Governance3%
- Pega NLP (Natural Language Processing)8%
- Pega Process AI8%
- 1
What is the primary function of the 'Trend' report in the Adaptive Decision Manager monitoring section?
Show answer details
Correct answer: A
The Trend report plots metrics like AUC and Success Rate over a timeline. This allows Data Scientists to see if a model is degrading, improving, or stabilizing over weeks or months.
- 2
In Pega Prediction Studio, you are using the wizard to build a new Predictive Model. Which step allows you to define the behavior you want to predict (e.g., 'Churn' vs 'Loyal')?
Show answer details
Correct answer: C
The Outcome Definition step is where you map the specific values in your training data (e.g., '1' or 'Yes') to the target behavior (e.g., 'Churn') that the model should learn to predict.
- 3
When importing a PMML model into Pega Prediction Studio, which version of PMML is generally recommended to ensure maximum compatibility with Pega's execution engine?
Show answer details
Correct answer: B
Pega Platform supports PMML standards, with the most robust support typically found for versions 4.x. Older versions (2.x, 3.x) are often deprecated or less supported.
- 4
You have trained a predictive model in Prediction Studio. The 'Validation' report shows a significant divergence between the 'Training' set performance (AUC 0.95) and the 'Validation' set performance (AUC 0.60). What is this phenomenon called?
Show answer details
Correct answer: B
Overfitting occurs when a model learns the training data (noise and all) too well but fails to generalize to new, unseen data (validation set). A large gap between training and validation scores is the classic indicator.
- 5
A Data Scientist wants to use an advanced Gradient Boosting model developed in H2O.ai driverless AI within a Pega Decision Strategy. Which file format should they export from H2O to import into Pega?
Show answer details
Correct answer: C
Pega supports H2O integration primarily via the MOJO format. MOJOs are optimized for deployment and scoring in real-time environments like Pega CDH.
- 6
A large retail bank is implementing Pega Customer Decision Hub (CDH) to transition from batch-based marketing to real-time interactions. They have a requirement to ensure that offers are arbitrated based on the customer's current financial situation calculated in real-time, rather than pre-calculated nightly scores. Which architectural component in the Next-Best-Action paradigm is specifically responsible for this real-time arbitration logic?
Show answer details
Correct answer: A
In the Pega CDH architecture, the 'Brain' represents the decisioning logic where Decision Strategies run. These strategies combine propensity scores, context, and business rules to arbitrate and select the Next-Best-Action in real-time.
- 7
A Data Scientist needs to create a new predictive model to estimate the likelihood of a customer churning. This model will be trained on historical data CSV files. Which workspace within the Pega Platform provides the dedicated environment for the entire lifecycle of this model, from creation to monitoring?
Show answer details
Correct answer: C
Prediction Studio is the dedicated workspace for data scientists to build, train, and monitor predictive and adaptive models, as well as manage text analytics and third-party model integrations.
- 8
You are explaining the concept of 'Predictor Power' in Pega's Adaptive Decision Manager (ADM) to a business stakeholder. Which statement accurately describes how ADM calculates this metric?
Show answer details
Correct answer: C
Predictor Power in Pega is a metric (typically Gini derived from AUC) that quantifies the ability of a specific predictor to differentiate between positive and negative outcomes. It ranges from 0 (random) to 100 (perfect prediction).
- 9
A telecom company launches a new 'Unlimited 5G' offer. They have configured an Adaptive Model to learn which customers are most likely to accept it. Since this is a brand new offer, there is no historical data. How does the Adaptive Decision Manager (ADM) handle the propensity calculation for the first few customers?
Show answer details
Correct answer: D
ADM is designed to handle the 'Cold Start' problem. Initially, when evidence is zero, it assigns a starting propensity (often 0.5 or a configured prior) and updates the model bins immediately with every new response, allowing the model to learn rapidly from the very first interaction.
- 10
While monitoring an Adaptive Model in Prediction Studio, you notice a predictor named 'Age' has a Predictor Power of 52%. However, another predictor 'Customer_ID' has a Predictor Power of 0%. What is the most likely reason ADM assigned 0% power to 'Customer_ID'?
Show answer details
Correct answer: B
ADM automatically evaluates predictors. Fields like unique IDs (high cardinality) are generally useless for prediction because they don't generalize. ADM detects this lack of correlation with the outcome and assigns a power of 0.
