Objective 4.1 Prepare Data Sets for Modeling CERTIFIED-DATA-SCIENCE-PRACTITIONER Practice Questions (Page 1)

Questions 1–5

1. 1foundation · easy

   In temporal splitting for time-series data, what is the typical approach to create training and test sets?

   • AInterleave training and test observations evenly across time
   • BUse the latest portion for training and the earliest for testing
   • CUse the earliest portion of the data for training and the latest portion for testing
   • DRandomly assign each observation to either training or test set
   Reveal Solution Discussion Select an answer first
2. 2application · easy

   A data science team is iterating on a model and wants to compare results across experiments. They are using a random split of their dataset. What should they do to ensure that the same data partition is used in every experiment?

   • ASet a fixed random seed before performing the split
   • BUse a different random seed for each experiment
   • CIncrease the size of the test set
   • DShuffle the data manually before each split
   Reveal Solution Discussion Select an answer first
4. 3foundation · easy

   In a typical supervised learning workflow, what is the primary purpose of the validation set?

   • ATo provide a final, unbiased estimate of model performance
   • BTo increase the size of the training data by combining with the training set
   • CTo tune hyperparameters and make model selection decisions
   • DTo train the model's weights and biases
   Reveal Solution Discussion Select an answer first
5. 4expert · hard

   A data scientist is developing a model and has used the test set multiple times to evaluate different versions of the model. They are concerned that the test set is no longer providing an unbiased estimate. What is the best course of action?

   • ACollect new data to create a fresh test set
   • BUse the validation set as the new test set
   • CUse the training set accuracy as the final performance metric
   • DContinue using the same test set, as it is the only one available
   Reveal Solution Discussion Select an answer first
6. 5foundation · easy

   Which of the following is a common source of data leakage in a machine learning pipeline?

   • ASplitting the data into training and test sets before any preprocessing
   • BRemoving duplicate rows from the dataset
   • CUsing a fixed random seed when splitting the data
   • DApplying feature scaling (e.g., standardization) using statistics computed from the entire dataset before splitting
   Reveal Solution Discussion Select an answer first