Objective 3.3 Train, Validate, and Test Data Subsets AIP-210 Practice Questions (Page 6)

Questions 26–28

1. 26foundation · easy

   What is look-ahead bias in time-series model evaluation?

   • AShuffling time-series data before splitting
   • BEvaluating a model on data it has already seen
   • CUsing future data to train a model that is then evaluated on past data
   • DUsing past data to predict future events
   Reveal Solution Discussion Select an answer first
2. 27foundation · easy

   Why is the validation subset used for hyperparameter tuning instead of the test subset?

   • ATo avoid biasing the final evaluation on unseen data
   • BThe validation subset is easier to compute
   • CThe validation subset is more representative of real-world data
   • DThe validation subset is larger than the test subset
   Reveal Solution Discussion Select an answer first
3. 28foundation · easy

   What is the primary purpose of the training subset in a machine learning workflow?

   • ATo evaluate the final model's performance on unseen data
   • BTo tune hyperparameters and select the best model
   • CTo fit the model's learnable parameters during training
   • DTo estimate the model's generalization error before deployment
   Reveal Solution Discussion Select an answer first