Objective 3.4 Evaluate the Model AIP-210 Practice Questions (Page 3)

Questions 11–15

1. 11expert · hard

   A data scientist fits a linear regression model and plots residuals versus fitted values. The plot shows a curved pattern. Which action is most appropriate to improve the model?

   • AAdd polynomial terms or use a nonlinear model
   • BApply a log transformation to the target variable
   • CRemove outliers from the dataset
   • DAdd interaction terms between features
   Reveal Solution Discussion Select an answer first
2. 12foundation · easy

   Which cross-validation technique partitions the dataset into k equal-sized folds and trains the model k times, each time using a different fold as the validation set?

   • ALeave-one-out cross-validation (LOOCV)
   • BStratified k-fold cross-validation
   • Ck-fold cross-validation
   • DHoldout validation
   Reveal Solution Discussion Select an answer first
4. 13foundation · easy

   When translating model evaluation metrics into business decisions, what is the most important factor to consider?

   • AThe training time of the model
   • BThe mathematical complexity of the model
   • CThe cost of false positives and false negatives in the specific business context
   • DThe number of features used in the model
   Reveal Solution Discussion Select an answer first
5. 14expert · hard

   A data scientist is building a model for a rare disease detection problem with only 200 positive samples and 10,000 negative samples. The scientist wants to evaluate the model's performance reliably. Which approach is most appropriate?

   • AUse leave-one-out cross-validation
   • BUse stratified k-fold cross-validation with a large number of folds (e.g., 10) and report the average AUC
   • CUse bootstrap sampling to create multiple training sets and average the test results
   • DUse a simple train-test split with 80% training and 20% testing
   Reveal Solution Discussion Select an answer first
6. 15application · medium

   A data scientist trains a decision tree model on a dataset with 10,000 samples. The training accuracy is 99.8%, but the validation accuracy is 82%. What is the most likely issue, and what is the best first step to address it?

   • AData leakage; split the data again using a different random seed
   • BUnderfitting; increase model complexity by adding more features
   • COverfitting; reduce model complexity by pruning the tree or limiting depth
   • DBias-variance tradeoff; increase the training dataset size
   Reveal Solution Discussion Select an answer first