PEGACPDS88V1 practice question 22 of 25

From two churn models with the similar performance, we chose the one the_____

Full set: 140 questions
Answer options
A: Highest churn rate
B: Highest number of predictors
C: Most evidence
D: Fewest number of predictors Correct answer
Correct Answer:
D. Fewest number of predictors
Explanation

This question adheres to the principle of parsimony, also known as Occam's Razor, a fundamental concept in machine learning and model building. When two predictive models yield similar performance (e.g., similar AUC or accuracy), the simpler model is preferred. A model with fewer predictors is considered simpler. This preference is because simpler models are less likely to be overfit to the training data, are more generalizable to new data, are easier to interpret and explain to business stakeholders, and are computationally less expensive to maintain and execute.

Why Incorrect

A. Highest churn rate: The model's purpose is to accurately predict churn, not to predict the highest possible rate. The churn rate is an outcome of the data, not a desirable characteristic of the model itself.
B. Highest number of predictors: A model with more predictors is more complex. Given similar performance, this complexity adds risks of overfitting and reduces interpretability without providing any predictive benefit.
C. Most evidence: This option is ambiguous. If "evidence" refers to the number of records used for training, it should be the same for both models for a fair comparison. It is not a standard criterion for model selection.

References

1. Pega Academy

Pega Certified Decisioning Consultant (8.8) Course

"Decisioning Models" Module

"Building a predictive model" Lesson: This lesson emphasizes the importance of selecting a concise set of powerful predictors. The methodology guides users to create models that are not only accurate but also transparent and robust

which is achieved by avoiding an unnecessary number of predictors that can lead to overfitting. The principle is to find the simplest model that adequately explains the variance.

2. Pega Documentation

"Prediction Studio"

Section: "Predictive model monitoring": While discussing model performance

the documentation implicitly supports parsimony. Models that are simpler (fewer predictors) are easier to monitor and manage over their lifecycle. When a simpler model performs as well as a complex one

it is the superior choice for production environments due to lower maintenance overhead.

3. Shmueli

G. (2010). To Explain or to Predict? Statistical Science

25(3)

289-310. This peer-reviewed article

while not Pega-specific

outlines foundational academic principles applied in platforms like Pega. It distinguishes between explanatory and predictive modeling

noting that for prediction

parsimonious models (simpler models) are often preferred as they tend to have better out-of-sample performance and are less likely to be overfit

a core concept in building reliable predictive models like those in Pega. (See Section 2.3 on Parsimony). DOI: https://doi.org/10.1214/10-STS330

Scroll to Top