Evaluate Model Performance SNOWPRO-ADVANCED-DATA-SCIENTIST Practice Questions (Page 4)

Questions 16–18

1. 16foundation · easy

   When comparing two classification models, which approach is most appropriate to determine if the difference in accuracy is statistically significant?

   • AUse a paired statistical test, such as McNemar's test, on the models' predictions.
   • BCompare the accuracy values directly and choose the higher one.
   • CUse a t-test on the accuracy values from a single train/test split.
   • DCompute the F1-score for both models and choose the one with the higher F1-score.
   Reveal Solution Discussion Select an answer first
2. 17application · medium

   A data scientist is evaluating a binary classifier for a disease screening test. The model achieves a recall of 0.95 and a precision of 0.20. The disease prevalence is 5%. Which conclusion is most appropriate?

   • AThe model is performing poorly because precision is low.
   • BThe model has a high false positive rate, which may be acceptable if follow-up tests are cheap.
   • CThe model is performing well because recall is high.
   • DThe model is overfitting because precision is low.
   Reveal Solution Discussion Select an answer first
3. 18foundation · easy

   What is the primary purpose of k-fold cross-validation?

   • ATo eliminate the need for a separate test set.
   • BTo ensure that the model is trained on all possible data combinations.
   • CTo provide a more reliable estimate of model performance than a single train/test split.
   • DTo increase the size of the training dataset.
   Reveal Solution Discussion Select an answer first