✅ Quiz M7.02

Question

We have a dataset with patient records from 10 different hospitals, and our goal is to predict whether a patient has a disease or not. Let’s also suppose that the classes (“disease” and “no-disease”) are imbalanced. Additionally, we suspect that each hospital’s data may have systematic biases due to factors like medical devices, policies, socioeconomic status of the patients, etc.

Which cross-validation strategy is the most suitable for assessing the model’s ability to make good predictions on patients from hospitals not seen during training?

• a) Group stratified k-fold cross-validation

• b) Group k-fold

• c) Stratified k-fold cross-validation

• d) Leave-one-out cross-validation

Select a single answer