Objective 3.3 Train, Validate, and Test Data Subsets AIP-210 Practice Questions (Page 3)

Questions 11–15

1. 11application · medium

   A data scientist is building a model to predict house prices. The dataset contains houses sold over the past 10 years. The scientist wants to evaluate the model's ability to predict prices for houses sold in the next year. Which splitting strategy is most appropriate?

   • ASplit chronologically so training uses houses sold before the last year and test uses houses sold in the last year
   • BUse all data for training and evaluate on a subset of the same data
   • CStratify by house size to ensure both sets have similar distributions
   • DRandomly split the data into 80% training and 20% test
   Reveal Solution Discussion Select an answer first
2. 12expert · hard

   A data science team is developing a model with a very small dataset of 500 samples. They need to tune hyperparameters and also provide an unbiased estimate of final model performance. Which approach best balances these two needs?

   • AUse a 70/15/15 split, tune on validation, and evaluate on test
   • BUse nested cross-validation: inner loop for hyperparameter tuning and outer loop for unbiased performance estimation
   • CUse 5-fold cross-validation for hyperparameter tuning, then train the final model on all data and report the cross-validation average as the final performance
   • DUse the entire dataset for training and report training accuracy
   Reveal Solution Discussion Select an answer first
4. 13foundation · easy

   What is a common split ratio for a moderately sized dataset?

   • A50% training, 25% validation, 25% test
   • B80% training, 10% validation, 10% test
   • C70% training, 15% validation, 15% test
   • D90% training, 5% validation, 5% test
   Reveal Solution Discussion Select an answer first
5. 14application · medium

   A data scientist has a dataset of 500,000 records and needs to split it into training, validation, and test sets. The dataset is large enough that a smaller validation and test set will still provide reliable estimates. Which split ratio is most appropriate for this scenario?

   • A90% training, 5% validation, 5% test
   • B60% training, 20% validation, 20% test
   • C98% training, 1% validation, 1% test
   • D50% training, 25% validation, 25% test
   Reveal Solution Discussion Select an answer first
6. 15application · medium

   A team is building a model to predict customer churn. The dataset contains customer features and a column indicating whether the customer churned. Before splitting, the team removes duplicate customer records. The team then splits the data into training and test sets. Which additional step is necessary to prevent data leakage?

   • AEnsure that the test set is larger than the training set
   • BEnsure that the training set contains only churned customers
   • CEnsure that no customer appears in both the training and test sets
   • DEnsure that the test set is stratified by customer age
   Reveal Solution Discussion Select an answer first