Certified Pega Data Scientist flashcards
Certified Pega Data Scientist · Build recall with a focused review session.
32 cards. Read the term and the definition here. The deck at the top of this page is optional.
- Rule Resolution
- The run-time algorithm Pega uses to pick one rule instance from many candidates, applying class specificity, ruleset stack order, circumstance, and availability in a defined precedence sequence.
- Circumstancing
- A technique that creates a more specific version of a rule that applies only when a property value, date range, or template condition matches, overriding the base rule at run time.
- Ruleset Stack
- The ordered list of rulesets in an application, where entries higher in the stack take precedence during rule resolution, enabling layered framework and application customization.
- Data Page
- A named clipboard structure that can cache data from an external source, with configurable scope such as thread or node and optional refresh thresholds for staleness.
- Access Group
- A run-time profile assigned to an operator that supplies the application, ruleset stack, portal, and default access role, determining what the user can see and do.
- pyWorkPage
- The clipboard page holding the primary case data for the current interaction; most UI rules, activities, and case processing default to this page.
- Standard Agent
- A background processing job defined on a node that runs scheduled system work, such as sending correspondence or processing queued items, without user intervention.
- Class Inheritance
- The mechanism by which an application class inherits rules from organization and framework parent classes, allowing a rule at a more specific class to override without editing the original.
- Predictive Model
- A machine learning artifact trained on historical case data that maps predictor properties to an outcome, producing a score and a predicted label for new cases.
- AUC
- Area under the receiver operating characteristic curve; a binary model metric where 0.5 means random prediction and values closer to 1.0 indicate better discrimination.
- Adaptive Model
- A model that continuously updates its scoring from incoming outcomes as cases are processed, rather than requiring a separate manual retraining cycle like a static model.
- Prediction Studio
- The Pega authoring environment where data scientists build, evaluate, and deploy predictive models and inspect predictor importance before use in case processing.
- Model Drift
- A degradation in model performance caused by changes in live data distributions relative to training data, detected by monitoring score and predictor distributions over time.
- Text Analytics
- The process of converting free-text case notes into structured predictors, such as extracted entities or sentiment scores, that machine learning models can consume.
- Holdout Set
- A portion of historical data withheld from model training and used to evaluate generalization, preventing overly optimistic performance estimates from overfitting.
- Predictor Importance
- A ranking of input properties by their contribution to a model's predictions, used to remove weak predictors and simplify the deployed model.
- Decision Strategy
- A directed graph of components such as decision, filter, prioritize, and output shapes that evaluates a page list and returns ranked results to the caller.
- Decision Table
- A rule that evaluates condition rows in order and returns the result of the first matching row, so row ordering directly determines the outcome.
- Arbitration
- The final selection step in next-best-action that applies business levers such as weights and caps to choose among eligible, suitable actions.
- Business Lever
- A configurable arbitration control, such as a weight that adjusts propensity or a cap limiting presentations per customer, used to steer offer selection.
- Propensity
- A model-produced score expressing the likelihood that a customer will accept or respond to an action, used to rank and prioritize propositions.
- Contact Policy
- A set of limits that suppresses actions when a customer has been contacted too often or too recently, applied before final arbitration selection.
- Data Flow
- A Pega rule that defines a pipeline of transformation shapes moving data between a source data set and a target data set, optionally with filtering.
- Champion-Challenger
- A testing approach that runs a control strategy alongside challenger variants on live or sampled traffic and compares lift to decide which to promote.
- Case Type
- The design-time definition of a case lifecycle, including its stages, processes, steps, data model, and optional alternate stages for exception handling.
- Alternate Stage
- A stage that a case enters on demand rather than automatically, often used for exceptions or parallel work, and can run alongside the primary stage.
- Assignment
- A case step that creates work for a user or work queue, requiring human action such as review or approval before the process continues.
- Service Level
- A rule defining goal, deadline, and passed-deadline behavior such as escalation or urgency increase, attached to a step, process, or case type.
- Work Queue
- A shared container of assignments that eligible users can pull from, with routing and skill rules determining which operators see the work.
- Subcase
- A child case created by and linked to a parent case, running its own lifecycle while sharing data and allowing the parent to wait on completion.
- Optimistic Locking
- The default case locking approach that detects conflicting saves when two users edit the same case and blocks the later save with a conflict message.
- Resolution Stage
- A stage configured to set the case status to Resolved, ending the case lifecycle and optionally triggering post-resolution processing or reporting.