PEGACPDS25V1 実際試験問題と解答 - 模擬練習 - 受験自信を増やす

Question No : 1
Which component can you use to integrate a prediction into a strategy?

A.Set Property
B.Prediction component
C.Decision Table
D.Data Page

正解:
Explanation:
The Prediction component is added to decision strategies to invoke predictions. It connects the strategy to Prediction Studio models and applies the prediction results in real time.

Question No : 2
Why would a data scientist choose a binary classifier over a scorecard?

A.More interpretability
B.Better for rule-based systems
C.Higher scalability for complex tasks
D.Lower processing speed

正解:
Explanation:
Binary classifiers scale well with complex, large datasets. They offer better performance and flexibility compared to static scorecards, which are rule-heavy and less adaptive.

Question No : 3
Which of the following are valid output types for Pega prediction models?

A.Text explanation
B.Propensity score
C.Recommendation ID
D.Outcome class

正解:
Explanation:
Prediction models generate outputs such as propensity scores (probabilities) and classified outcomes (e.g., churn/retain). These outputs guide strategy decisions.

Question No : 4
Which of the following can improve the accuracy of a predictive model? (Choose two)

A.Drop all categorical variables
B.Engineer new features
C.Remove noisy predictors
D.Set AUC to zero

正解:
Explanation:
Feature engineering and removing noisy predictors can significantly enhance model accuracy. These steps improve signal clarity and reduce confusion in the learning process.

Question No : 5
How is lift used in evaluating a prediction?

A.To reduce AUC
B.To normalize predictors
C.To compare model with random chance
D.To create outcome rules

正解:
Explanation:
Lift measures how much better the model performs compared to random selection. A higher lift indicates a stronger ability to identify positive outcomes.

Question No : 6
Why should you review a confusion matrix for your prediction?

A.To understand UI rules
B.To determine model scoring rules
C.To analyze true and false positives/negatives
D.To track deployment status

正解:
Explanation:
The confusion matrix summarizes model predictions and outcomes. It helps analyze the balance between true/false positives and negatives, aiding in tuning and threshold setting.

Question No : 7
What are some common causes for poor model performance? (Choose two)

A.Too many model thresholds
B.Overfitting
C.Data imbalance
D.Excess predictors

正解:
Explanation:
Poor performance can stem from overfitting (model learns noise) or class imbalance (skewed outcome distribution), leading to inaccurate or biased predictions.

Question No : 8
What does AUC indicate about a prediction model?

A.Model latency
B.Feature engineering strength
C.Discriminatory power
D.Execution speed

正解:
Explanation:
AUC (Area Under the Curve) measures how well the model separates positive and negative outcomes. A value closer to 1 indicates high predictive power.

Question No : 9
Which steps are involved in deploying a prediction? (Choose two)

A.Creating a new engagement policy
B.Mapping the prediction to a strategy
C.Configuring predictors manually
D.Connecting it to an outcome path

正解:
Explanation:
To deploy a prediction, it must be mapped into a decision strategy and linked to an outcome path. This allows the model to influence real-time action decisions.

Question No : 10
Which metric indicates the balance between precision and recall?

A.F1-score
B.Lift
C.Gini coefficient
D.AUC

正解:
Explanation:
The F1-score is the harmonic mean of precision and recall. It provides a balanced measure of model accuracy when false positives and false negatives are both costly.

Question No : 11
What does the term “training data” refer to in prediction models?

A.UI interaction logs
B.Test environments
C.Historical labeled data
D.Predictor type templates

正解:
Explanation:
Training data consists of historical cases with known outcomes. This data is used by the model to learn patterns and make accurate future predictions.

Question No : 12
Which two elements must be mapped when setting up outcomes in a prediction? (Choose two)

A.Positive label
B.Source model ID
C.Prediction threshold
D.Negative label

正解:
Explanation:
Outcome mapping requires identifying which values from the input data represent positive and negative responses. This allows the model to learn the success/failure patterns.

Question No : 13
How can you validate the quality of a new prediction model?

A.Run adaptive tuning
B.Export the evidence
C.Evaluate ROC and AUC scores
D.Publish to CDH

正解:
Explanation:
ROC (Receiver Operating Characteristic) and AUC (Area Under Curve) scores are used to evaluate model quality. They indicate how well the model discriminates between classes.

Question No : 14
Which of the following can be used as a predictor in a prediction model? (Choose two)

A.Treatment ID
B.Age
C.Last purchase amount
D.Case ID

正解:
Explanation:
Predictors are variables that correlate with outcomes. Attributes like age or purchase amount are common predictors that help inform the model’s decision-making.

Question No : 15
Why would a data scientist use a scorecard over a binary classifier?

A.To build rule-based thresholds
B.To analyze model history
C.To remove predictors
D.To deploy adaptive models

正解:
Explanation:
Scorecards use rule-based scoring logic, making them interpretable and easier to validate. They're useful for regulated environments needing transparent decision logic.

Pegasystems PEGACPDS25V1 問題練習