QUESTION 1 (Classification model evaluation)
15 marks
Assume that two individuals offer to sell you their predictive models M. and M2. The
confusion matrices produced by each model are as follows.
Predicted Predicted
True
False
5
95
Actually
True
Actually
False
10
90
Performance of M.
Predicted Predicted
True False
85
15
Actually
True
Actually
False
95
5
Performance of M2
1. What is the accuracy of each model?
2. Assuming that precision is of paramount importance in your application, which of
the two models would you buy? Why?
3. Assuming that the cost of labeling as True something that is actually False far
exceeds the cost of labeling as False something that is actually True, which of the
two models would you buy? Why?