A classifier of fake banknotes is being tested to see how reliable it can detect real banknotes from fake ones. Three confusion matrices representing 3 Days of testing was made
Day 1
n = 100
Predicted No Predicted Yes
Actual No - 15 35
Actual Yes - 8 42
Day 2
n = 100
Predicted No Predicted Yes
Actual No - 22 28
Actual Yes - 2 48
Day 3
n = 100
Predicted No Predicted Yes
Actual No - 43 7
Actual Yes - 16 34
1)Which day has the worst CSI?
A) Day 1
B) Day 2
C) Day 3
D) Insufficient Information. Cannot be determined
2)The overall F-score of the test is around? 4 decimals places
3) If the target of the study is to have a total CSI above 60% and an F-Score of above 70%, does the classifier pass? True or False
1)
If CSI is sensitivity, then:
"CSI=\\frac{TP}{TP+FN}"
where TP is true positive, FN is false negative.
"CSI_1=\\frac{42}{42+8}=0.84"
"CSI_2=\\frac{48}{48+2}=0.96"
"CSI_3=\\frac{34}{34+16}=0.68"
Answer: A) Day 3 has the worst CSI.
2)
F-score:
"F=\\frac{2TP}{2TP+FP+FN}"
where FP is false positive
"F=\\frac{2(42+48+34)}{2(42+48+34)+(35+28+7)+(8+2+16)}=0.72"
3)
True. The classifier passes, because F-score of the test is more than the target of the study.
Comments
Leave a comment