Hands-On Exercise: Identify Training and Test Data and Create an ML Model CT-AI Practice Questions (Page 4)

Questions 16–20

1. 16foundation · easy

   What is the purpose of evaluating a trained model on the test dataset?

   • ATo select the best hyperparameters for the model
   • BTo estimate how well the model will perform on new, unseen data
   • CTo visualize the training data distribution
   • DTo improve the model's accuracy by further training on test data
   Reveal Solution Discussion Select an answer first
2. 17application · medium

   A team has built a binary classifier to predict loan default. They evaluate it on a test set and obtain the following confusion matrix: True Positives (default predicted as default) = 50, False Negatives (default predicted as no-default) = 50, False Positives (no-default predicted as default) = 100, True Negatives (no-default predicted as no-default) = 800. What is the precision of this model?

   • A0.833
   • B0.500
   • C0.100
   • D0.333
   Reveal Solution Discussion Select an answer first
4. 18foundation · easy

   What is the first step in creating a simple ML model from a dataset?

   • AEvaluate the model on the training set
   • BDeploy the model to production
   • CSplit the dataset into training and test sets
   • DTrain the model on the entire dataset
   Reveal Solution Discussion Select an answer first
5. 19application · medium

   A machine learning engineer is evaluating two models for a binary classification task. Model A has an AUC of 0.90, and Model B has an AUC of 0.85. Which of the following is the most accurate interpretation of these results?

   • AModel A is better at ranking positive instances higher than negative instances.
   • BModel A has a lower false positive rate than Model B.
   • CModel A has higher accuracy than Model B.
   • DModel B is less overfit than Model A.
   Reveal Solution Discussion Select an answer first
6. 20application · medium

   A machine learning engineer is developing a model to predict patient readmission within 30 days. She has a dataset of 50,000 patient records. She wants to ensure her model's performance estimate is reliable and not overly optimistic. Which of the following approaches is the most appropriate for this purpose?

   • AUse k-fold cross-validation on the training data and report the average performance.
   • BTrain the model on the entire dataset and report the training accuracy.
   • CUse a single random train/test split and report the test accuracy without repeating the process.
   • DSplit the data into a training set and a test set, but only evaluate on the training set.
   Reveal Solution Discussion Select an answer first