Suppose you are trying to train and test a Logistic Regression Classifier (binary) where 90% of the observations in the training and test set belong to the positive class (y=1). After training your classifier under these conditions, it will probably end up predicting y = 1 for all examples in the test set. Considering that the abovementioned happens:

1) What plausible metrics will you expect?

*note you must select all the answers that apple. Also, you must explain why and how the selected option(s) are plausible to happen.

a. The Missclassification Error (ME) will be small (e.g. less than 20%)

b. The Accuracy will be high (e.g. higher than 80%)

c. The F-Score will be small (e.g. less than 0.5)

d. The F-Score will be high (e.g. higher than 0.5)

e. Recall will be high, and Precision will be small

f. Precision will be high, and Recall will be small

2) What happens to the Misclassification Error (ME) if we know train the classifier with 90% of the observations belonging to the negative class instead (y=0)?

*select the correct answer and explain

a. The ME will be smaller

b. The ME will be bigger

c. The ME will remain roughly the same

3) Explain why training under any of the previous circumstances produces a Logistic Regression Classifier predicting y = 0 or y = 1 for almost all examples in the test set.