Factors Involved in ML Algorithm Selection CT-AI Practice Questions (Page 6)

Questions 26–29

1. 26application · medium

   An e-commerce company wants to segment its customers into distinct groups based on purchasing behavior, browsing history, and demographics. They have 2 million customer records with numerical and categorical features. The marketing team needs the segments to be interpretable for campaign design. They have a standard cloud cluster with moderate compute. Which approach is most appropriate?

   • AHierarchical clustering with Ward's linkage
   • BDBSCAN with default parameters
   • CK-means clustering with standardized features
   • DApriori association rule mining
   Reveal Solution Discussion Select an answer first
2. 27foundation · easy

   Which of the following is a computational constraint that should be considered when selecting an ML algorithm?

   • AThe business goal
   • BThe interpretability of the model
   • CScalability to large datasets
   • DThe data type
   Reveal Solution Discussion Select an answer first
4. 28expert · hard

   A data science team is building a model to predict customer churn. They have a dataset with 50,000 records and 200 features. They have tried a single decision tree and noticed high variance in performance across different training/test splits. They need to improve generalization while keeping the model reasonably interpretable. Which approach is most appropriate?

   • AUse a random forest with a limited number of trees
   • BUse a single decision tree with increased depth
   • CUse logistic regression without regularization
   • DUse a support vector machine with RBF kernel
   Reveal Solution Discussion Select an answer first
5. 29application · medium

   A credit union needs to approve or deny loan applications. Regulatory requirements mandate that the model must provide a clear explanation for each decision. The dataset has 100,000 records with 50 features. The team has tested several models and found that a gradient boosting model achieves 95% accuracy, while a decision tree achieves 88% accuracy. Which model should they deploy?

   • AA random forest, as a compromise
   • BDecision tree, because interpretability is a regulatory requirement
   • CGradient boosting with SHAP explanations
   • DGradient boosting, because higher accuracy is more important
   Reveal Solution Discussion Select an answer first