29165fa9
 David Mayerich
added ROC curve c...
|
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
|
function [TPR, FPR, AUC] = stimROC(C, T)
%build an ROC curve
% C - class labels as an array of binary values (1 = true positive)
% T - threshold used for classification
%sort the thresholds in descending order and get the indices
[~, I] = sort(T, 'descend');
%sort the class labels in the same order as the thresholds
Cs = C(I);
%calculate the number of measurements
M = size(C, 2);
%calculate the number of positives
P = nnz(C);
%calculate the number of negatives
N = M - P;
%if all examples are positives or negatives, return a perfect score?
if P == M
error('ERROR: no positive observations');
end
if P == 0
error('ERROR: no negative observations');
end
%allocate space for the ROC curve
TPR = zeros(1, M);
FPR = zeros(1, M);
%calculate the number of inflection points
ip = 0;
for i = 2:M
if Cs(i) ~= Cs(i-1)
ip = ip + 1;
end
end
%initialize the true and false positive rates to zero
TP = 0;
FP = 0;
for i = 1:M
if Cs(i) == 1
TP = TP + 1;
else
FP = FP + 1;
end
TPR(i) = TP / P;
FPR(i) = FP / N;
end
%calculate the area under the ROC curve
AUC = 0;
for i = 2:M
w = FPR(i) - FPR(i-1);
h = TPR(i);
AUC = AUC + w * h;
end
|
