def truePositiveRate(self,classIndex:int):
correct=total=0
for j in range(self.m_NumClasses):
if j == classIndex:
correct+=self.m_ConfusionMatrix[classIndex][j]
total+=self.m_ConfusionMatrix[classIndex][j]
return Utils.division(correct, total)
def precision(self,classIndex:int):
correct=total=0
for i in range(self.m_NumClasses):
if i == classIndex:
correct+=self.m_ConfusionMatrix[i][classIndex]
total+=self.m_ConfusionMatrix[i][classIndex]
return Utils.division(correct, total)
def falsePositiveRate(self,classIndex:int):
incorrect=total=0
for i in range(self.m_NumClasses):
if i != classIndex:
for j in range(self.m_NumClasses):
if j == classIndex:
incorrect+=self.m_ConfusionMatrix[i][j]
total+=self.m_ConfusionMatrix[i][j]
return Utils.division(incorrect, total)
def matthewsCorrelationCoefficient(self,classIndex:int):
numTP=self.numTruePositives(classIndex)
numTN=self.numTrueNegatives(classIndex)
numFP=self.numFalsePositives(classIndex)
numFN=self.numFalseNegatives(classIndex)
n=numTP*numTN-numFP*numFN
d=(numTP+numFP)*(numTP+numFN)*(numTN+numFP)*(numTN+numFN)
d=math.sqrt(d)
return Utils.division(n, d)
def weightedAreaUnderPRC(self):
classCounts = [0] * self.m_NumClasses
classCountSum = 0
for i in range(self.m_NumClasses):
for j in range(self.m_NumClasses):
classCounts[i] += self.m_ConfusionMatrix[i][j]
classCountSum += classCounts[i]
auprcTotal = 0
for i in range(self.m_NumClasses):
temp = self.areaUnderPRC(i)
if classCounts[i] > 0:
auprcTotal += temp * classCounts[i]
return Utils.division(auprcTotal, classCountSum)
def weightedMatthewsCorrelation(self):
classCounts = [0] * self.m_NumClasses
classCountSum = 0
for i in range(self.m_NumClasses):
for j in range(self.m_NumClasses):
classCounts[i] += self.m_ConfusionMatrix[i][j]
classCountSum += classCounts[i]
mccTotal = 0
for i in range(self.m_NumClasses):
temp = self.matthewsCorrelationCoefficient(i)
if classCounts[i] > 0:
mccTotal += temp * classCounts[i]
return Utils.division(mccTotal, classCountSum)
def weightedFMeasure(self):
classCounts = [0] * self.m_NumClasses
classCountSum = 0
for i in range(self.m_NumClasses):
for j in range(self.m_NumClasses):
classCounts[i] += self.m_ConfusionMatrix[i][j]
classCountSum += classCounts[i]
fMeasureTotal = 0
for i in range(self.m_NumClasses):
temp = self.fMeasure(i)
if classCounts[i] > 0:
fMeasureTotal += temp * classCounts[i]
return Utils.division(fMeasureTotal, classCountSum)
def weightedPrecision(self):
classCounts=[0]*self.m_NumClasses
classCountSum=0
for i in range(self.m_NumClasses):
for j in range(self.m_NumClasses):
classCounts[i]+=self.m_ConfusionMatrix[i][j]
classCountSum+=classCounts[i]
precisionTotal=0
for i in range(self.m_NumClasses):
temp=self.precision(i)
if classCounts[i]>0:
precisionTotal+=temp*classCounts[i]
return Utils.division(precisionTotal, classCountSum)
def weightedTruePositiveRate(self):
classCounts=[0]*self.m_NumClasses
classCountSum=0
for i in range(self.m_NumClasses):
for j in range(self.m_NumClasses):
classCounts[i]+=self.m_ConfusionMatrix[i][j]
classCountSum+=classCounts[i]
truePosTotal=0
for i in range(self.m_NumClasses):
temp=self.truePositiveRate(i)
if classCounts[i]>0:
truePosTotal+=temp*classCounts[i]
return Utils.division(truePosTotal, classCountSum)
