Suppose we have 100 individuals, some of whom have type 2 diabetes. We use a cla
ID: 3046246 • Letter: S
Question
Suppose we have 100 individuals, some of whom have type 2 diabetes. We use a classification model based on health and lifestyle profiles of the individuals to estimate the probability that each individual is diabetic. If we use a 0.5 probability cut-off, the model predicts that 60 individuals are diabetic, of whom 54 individuals actually are diabetic, while of the other 40 individuals, only 5 are diabetic. Which of the following are possible results that could result from increasing the cut-off probability to 0.6? (Select all that apply.)
A) 47 actual diabetics out of 49 predicted diabetics; 12 actual diabetics out of 51 predicted non-diabetics.
B) 50 actual diabetics out of 53 predicted diabetics; 2 actual diabetics out of 47 predicted non-diabetics.
C) 55 actual diabetics out of 67 predicted diabetics; 6 actual diabetics out of 33 predicted non-diabetics.
D) 58 actual diabetics out of 69 predicted diabetics; 1 actual diabetic out of 31 predicted non-diabetics.
Explanation / Answer
SUm of actual diabetics is 54+5=59
THis is possible only in options A and D.
Out of these options, increasing probability cutoff will decrease the number of predicted diabetics
SO , A is the correct option