Objective 7.1 Report Findings CERTIFIED-DATA-SCIENCE-PRACTITIONER Practice Questions (Page 13)

Questions 61–64

1. 61expert · hard · select all that apply

   A data science team is reviewing a colleague's model training code before deployment. The code works but is difficult to understand. The team wants to improve maintainability. Which practices are most effective for making the code understandable? (Select all that apply.)

   • ARemove all comments to keep the code short and clean.
   • BWrite a single long script with no functions to avoid indirection.
   • CUse descriptive variable names that reflect the data they hold.
   • DAdd comments explaining the purpose of each major section and any non-obvious decisions.
   • EDocument the overall workflow in a README or docstring, including data sources and expected outputs.
   Check answers Discussion Select an answer first
2. 62foundation · easy

   What is a potential drawback of using a black-box model when explainability is required?

   • AIt requires more feature engineering
   • BIt is always less accurate than interpretable models
   • CIt cannot handle large datasets
   • DIt may be difficult to justify its decisions to stakeholders
   Reveal Solution Discussion Select an answer first
4. 63foundation · easy

   What is the primary purpose of LIME (Local Interpretable Model-agnostic Explanations)?

   • ATo reduce the model's complexity
   • BTo improve the model's accuracy
   • CTo explain individual predictions by approximating the model locally with an interpretable surrogate
   • DTo explain the global behavior of the model across all predictions
   Reveal Solution Discussion Select an answer first
5. 64expert · hard

   A data science team is starting a new project to predict equipment failure. The team has historical sensor data but is unsure whether to spend time on feature engineering or model selection first. The project has a tight deadline. Which approach best reflects sound ML workflow design?

   • ASpend most of the time on complex feature engineering before training any model.
   • BStart with a simple baseline model using raw features, then iterate on feature engineering and model complexity based on baseline performance.
   • CSelect the most complex model available and train it immediately.
   • DSkip model training and use a rule-based system based on domain knowledge.
   Reveal Solution Discussion Select an answer first