Objective 3.3 Train, Validate, and Test Data Subsets AIP-210 Practice Questions (Page 2)

Questions 6–10

1. 6foundation · easy

   During model training, which subset is used to update the model's parameters?

   • AValidation subset
   • BTraining subset
   • CTest subset
   • DAll subsets equally
   Reveal Solution Discussion Select an answer first
2. 7application · medium

   A team is building a model to classify customer reviews as positive or negative. The reviews were collected over several years and the language used has changed over time. The team wants to evaluate how well the model will perform on future reviews. Which splitting approach is most appropriate?

   • AUse all reviews for training and evaluate on a sample of the same reviews
   • BStratify by review length to ensure both subsets have similar text lengths
   • CRandomly shuffle all reviews and split into 80% training and 20% test
   • DSplit by time so training uses older reviews and test uses the most recent reviews
   Reveal Solution Discussion Select an answer first
4. 8foundation · easy

   What is the main purpose of stratified splitting when creating training and test subsets?

   • ATo maintain the original class distribution across subsets
   • BTo randomly shuffle the data to avoid order bias
   • CTo ensure each subset has the same number of samples
   • DTo increase the size of the training subset
   Reveal Solution Discussion Select an answer first
5. 9expert · hard

   A team is building a model to predict hospital readmission risk. The dataset includes patient records and a feature engineered from the full dataset: the average length of stay for all patients with the same primary diagnosis. The team splits the data into training and test sets, trains the model, and achieves 98% accuracy on the test set, which seems unrealistically high. What is the most likely cause of this inflated performance?

   • AThe engineered feature was computed using the entire dataset before splitting, leaking test information into training
   • BThe model is overfitting the training data due to too many features
   • CThe test set is too small, causing high variance in the accuracy estimate
   • DThe model is using a validation set during training, which biases the test results
   Reveal Solution Discussion Select an answer first
6. 10expert · hard

   A team is working with a small, highly imbalanced dataset where the positive class represents only 5% of samples. They need to tune hyperparameters and want a reliable performance estimate. They plan to use k-fold cross-validation. Which modification is most important to ensure the cross-validation estimate is reliable?

   • AUse stratified k-fold cross-validation so each fold preserves the 5% positive class ratio
   • BUse a single validation split instead of cross-validation to save computation
   • CIncrease the number of folds to 20 to make each fold smaller
   • DShuffle the data randomly before each fold split without stratification
   Reveal Solution Discussion Select an answer first