Plot ROC CURVE

[Salvatore Esposito]

Hi,

I implemented, in Matlab, a Bayes Detector, my code return me 2 vectors (100
x1):

TargetDecision: Bayes Detector Decision (my code) (1 for target presence, 0
for target absence)
TargetTrue: real target presence in my observation .

Because I want to plot a ROC curve, I put it in for cycle in this way :

%%%%%%%%
for 1=1:NumberExtraction

[TargetDecision TargetTrue]=BayesDetector(,...,....,)

TruePositive = length(find( TargetDecision==1)&(TargetTrue==1))
FalseNegative=length(find( TargetDecision==0)&(TargetTrue==1))
FalsePositive=length(find( TargetDecision==1)&(TargetTrue==0))
TrueNegative=length(find( TargetDecision==0)&(TargetTrue==0))

TruePositiveRate(i,1)=TruePositive / [TruePositive + FalseNegative]
FalsePositiveRate(i,1)=FalsePositive / [FalsePositive + TrueNegative]
end

plot(FalsePositiveRate, TruePositiveRate)

but it dosen't work. I don't have a correct RoC curve. Why? Is there a
conceptual mistake?

Thanks a lot for your help
Salvatore

[marwa hadhoud]

you need to sort the FPR in ascending order and its corresponding TPR before plotting the ROC curve

"Salvatore Esposito" <salvatore...@alice.it> wrote in message <469e6810$0$4792$4faf...@reader4.news.tin.it>...