Certified Pega Data Scientist Free Sample Questions

The Prediction Studio portal

The Decision Strategy within the 'Brain'

The Interaction History data store

The Job Scheduler for campaign execution

Dev Studio

App Studio

Prediction Studio

Admin Studio

It measures the correlation between a single predictor and the outcome, expressed as a percentage (0-100) based on the Area Under the Curve (AUC).

It represents the number of times a predictor was used in a decision strategy divided by the total number of decisions.

It is the percentage of missing values found in the predictor data stream over the last 24 hours.

It calculates the speed at which the predictor data is retrieved from the customer database.

It returns a hardcoded value of 0.5 for all customers until at least 100 responses are recorded.

It uses a 'Cold Start' mechanism, often starting with a base propensity (e.g., 0.5) and rapidly adjusting as the first pieces of evidence arrive.

It blocks the offer from being presented until a Data Scientist manually uploads a seed file.

It uses the average propensity of all other active offers in the system.

The 'Customer_ID' field contains too many unique values (high cardinality), making it statistically insignificant for predicting general behavior.

The 'Customer_ID' is a numeric field, and ADM only supports symbolic predictors.

The 'Customer_ID' predictor was manually disabled in the strategy.

The system cannot process text strings longer than 10 characters.

The predictor has low performance but high frequency of use.

The predictor is highly predictive (high AUC) and is frequently present in the data (high support).

The predictor is redundant and correlated with another predictor.

The predictor has a high error rate and should be removed.

ADM groups correlated predictors and only uses the one with the highest performance from the group for the active model scoring.

ADM averages the values of both predictors into a new synthetic predictor.

ADM deletes both predictors and requests the user to choose one manually.

It does not handle this; the Data Scientist must remove one predictor from the input definition.

The model is predicting perfectly.

The model has no predictive power and is no better than a random guess.

The model is overfitted to the training data.

The model has encountered a technical error during training.

The 'Update Frequency' setting in the ADM Service configuration (e.g., every 50 responses or every hour).

Models are updated instantly after every single click.

The 'Model Update Frequency' is configured on the Model rule form, but the actual update happens periodically based on the ADM service snapshot interval.

ADM models only update when a manual 'Retrain' button is clicked in Prediction Studio.

Add 'Rejected' as a negative outcome in the Adaptive Model rule and potentially assign it a higher weight if the version supports weighted outcomes.

You cannot; ADM only supports one type of positive and one type of negative outcome.

Create a separate model for Rejections and subtract its score from the Acceptance model.

Use a Predictive Model instead, as Adaptive Models cannot handle multiple outcome types.