✅ Quiz#


With a same dataset, feature importance might differs if:

  • a) we use two different models

  • b) we use two different train/test split with a same model

  • c) we use a same model with a different set of hyper-parameters

  • d) we use a same model with the same set of hyper-parameters but a different random_state


In linear model, the feature importance:

  • a) might be infer from the coefficients

  • b) might be infer by importance_permutation

  • c) need a regularization to infer the importance

  • d) is a built-in attribute


If two feature are the same (thus correlated)

  • a) their feature importance will be the same

  • b) their feature importance will be divided by 2

  • c) only one will receive all the feature importance, the second one will be 0

  • d) it depends


The feature importance provided by the scikit-learn random forest:

  • a) has bias for categorical feature

  • b) has bias for continuous (high cardinality) feature

  • c) is independent from the train/test split

  • d) is independent from the hyper-parameters


To evaluate the feature importance for a specific model, one could:

  • a) drop a column and compare the score

  • b) shuffle a column and compare the score

  • c) put all column to 0 and compare the score

  • d) change a column value to random number and compare the score