def prune(self):
if not self.m_isLeaf:
for i in range(len(self.m_sons)):
self.son(i).prune()
indexOfLargestBranch = self.localModel().distribution().maxBag()
if self.m_subtreeRaising:
errorsLargestBranch = self.son(
indexOfLargestBranch).getEstimatedErrorsForBranch(
self.m_train)
else:
errorsLargestBranch = float("inf")
errorsLeaf = self.getEstimatedErrorsForDistribution(
self.localModel().distribution())
errorsTree = self.getEstimatedErrors()
if (Utils.gr(errorsTree+0.1, errorsLeaf) or Utils.equal(errorsTree+0.1, errorsLeaf)) and\
(Utils.gr(errorsLargestBranch+0.1, errorsLeaf) or Utils.equal(errorsLargestBranch+0.1, errorsLeaf)):
self.m_sons = None
self.m_isLeaf = True
self.m_localModel = NoSplit(self.localModel().distribution())
return
if Utils.gr(errorsTree + 0.1, errorsLargestBranch) or Utils.equal(
errorsTree + 0.1, errorsLargestBranch):
largestBranch = self.son(indexOfLargestBranch)
self.m_sons = largestBranch.m_sons
self.m_localModel = largestBranch.localModel()
self.m_isLeaf = largestBranch.m_isLeaf
self.newDistribution(self.m_train)
self.prune()
def split(self,data:Instances)->List[Instances]:
subsetSize=[0]*self.m_numSubsets
for inst in data:
subset=self.whichSubset(inst)
if subset > -1:
subsetSize[subset]+=1
else:
weights=self.weights(inst)
for j in range(self.m_numSubsets):
if Utils.gr(weights[j], 0):
subsetSize[j]+=1
instances=[] #type:List[Instances]
for j in range(self.m_numSubsets):
instances.append(Instances(data,subsetSize[j]))
for inst in data:
subset=self.whichSubset(inst)
if subset > -1:
instances[subset].add(inst)
else:
weights=self.weights(inst)
for j in range(self.m_numSubsets):
if Utils.gr(weights[j], 0):
instances[j].add(inst)
instances[j].lastInstance().setWeight(float(weights[j]*inst.weight()))
return instances
def computeAverageClassValues(self):
avgClassValues = [[]
for i in range(self.getInputFormat().numAttributes())
]
self.m_Indices = [[]
for i in range(self.getInputFormat().numAttributes())
]
for j in range(self.getInputFormat().numAttributes()):
att = self.getInputFormat().attribute(j)
if att.isNominal():
avgClassValues[j] = [0] * att.numValues()
counts = [0] * att.numValues()
for i in range(self.getInputFormat().numInstances()):
instance = self.getInputFormat().instance(i)
if not instance.classIsMissing(
) and not instance.isMissing(j):
counts[int(instance.value(j))] += instance.weight()
avgClassValues[j][int(instance.value(
j))] += instance.weight() * instance.weight()
sums = sum(avgClassValues[j])
totalCounts = sum(counts)
if Utils.gr(totalCounts, 0):
for k in range(att.numValues()):
if Utils.gr(counts[k], 0):
avgClassValues[j][k] /= counts[k]
else:
avgClassValues[j][k] = sums / totalCounts
self.m_Indices[j] = Utils.sortDouble(avgClassValues[j])
def selectModel(self, data: Instances, test: Instances = None):
if test is not None:
return self.selectModel(data)
multiVal = True
averageInfoGain = validModels = 0
checkDistribution = Distribution(data)
noSplitModel = NoSplit(checkDistribution)
if Utils.gr(2*self.m_minNoObj, checkDistribution.total()) or \
Utils.equal(checkDistribution.total(), checkDistribution.perClass(checkDistribution.maxClass())):
return noSplitModel
if self.m_allData is not None:
for attr in data.enumerateAttributes():
if attr.isNumeric() or Utils.gr(
0.3 * self.m_allData.numInstances(), attr.numValues()):
multiVal = False
break
currentModel = [None] * data.numAttributes() #type:List[C45Split]
sumOfWeights = data.sumOfWeight()
for i in range(data.numAttributes()):
if i != data.classIndex():
currentModel[i] = C45Split(i, self.m_minNoObj, sumOfWeights,
self.m_useMDLcorrection)
currentModel[i].buildClassifer(data)
if currentModel[i].checkModel():
if self.m_allData is not None:
if data.attribute(i).isNumeric() or \
(multiVal or Utils.gr(0.3*self.m_allData.numInstances(), data.attribute(i).numValues())):
averageInfoGain = averageInfoGain + currentModel[
i].infoGain()
validModels += 1
else:
averageInfoGain = averageInfoGain + currentModel[
i].infoGain()
validModels += 1
else:
currentModel[i] = None
if validModels == 0:
return noSplitModel
averageInfoGain = averageInfoGain / validModels
minResult = 0
for i in range(data.numAttributes()):
if i != data.classIndex() and currentModel[i].checkModel():
if currentModel[i].infoGain() >= averageInfoGain-1e-3 and\
Utils.gr(currentModel[i].gainRatio(), minResult):
bestModel = currentModel[i]
minResult = currentModel[i].gainRatio()
if Utils.equal(minResult, 0):
return noSplitModel
bestModel.distribution().addInstWithUnknown(data, bestModel.attIndex())
if self.m_allData is not None and not self.m_doNotMakeSplitPointActualValue:
bestModel.setSplitPoint(self.m_allData)
return bestModel
def splitCritValue(self,
bags: Distribution,
totalNoInst: float = None,
numerator: float = None):
if totalNoInst is None and numerator is None:
numerator = self.oldEnt(bags) - self.newEnt(bags)
if Utils.equal(numerator, 0):
return float('inf')
denumerator = self.splitEnt(bags)
if Utils.equal(denumerator, 0):
return float('inf')
return denumerator / numerator
elif numerator is None:
res = 0
noUnkown = totalNoInst - bags.total()
if Utils.gr(bags.total(), 0):
for i in range(bags.numBags()):
res = res - self.lnFunc(bags.perBag(i))
res = res - self.lnFunc(noUnkown)
res = res + self.lnFunc(totalNoInst)
return res / math.log(2)
else:
denumerator = self.splitEnt(bags, totalNoInst)
if Utils.equal(denumerator, 0):
return 0
denumerator /= totalNoInst
return numerator / denumerator
def buildClassifier(self, instances: Instances):
self.getCapabilities().testWithFail(instances)
sumOfWeights = 0
self.m_Class = instances.classAttribute()
self.m_ClassValue = 0
attrType = instances.classAttribute().type()
if attrType == Attribute.NUMERIC:
self.m_Counts = None
elif attrType == Attribute.NOMINAL:
self.m_Counts = []
for i in range(instances.numClasses()):
self.m_Counts.append(1)
sumOfWeights = instances.numClasses()
for instance in instances:
classValue = instance.classValue()
if not Utils.isMissingValue(classValue):
if instances.classAttribute().isNominal():
self.m_Counts[classValue] += instance.weight()
else:
self.m_ClassValue += instance.weight() * classValue
sumOfWeights += instance.weight()
if instances.classAttribute().isNumeric():
if Utils.gr(sumOfWeights, 0):
self.m_ClassValue /= sumOfWeights
else:
self.m_ClassValue = Utils.maxIndex(self.m_Counts)
Utils.normalize(self.m_Counts, sumOfWeights)
def maxClass(self, index: int = None):
maxCount = 0
maxIndex = 0
if index is None:
for i in range(len(self.m_perClass)):
if Utils.gr(self.m_perClass[i], maxCount):
maxCount = self.m_perClass[i]
maxIndex = i
return maxIndex
else:
if Utils.gr(self.m_perBag[index], 0):
for i in range(len(self.m_perClass)):
if Utils.gr(self.m_perClassPerBag[index][i], maxCount):
maxCount = self.m_perClassPerBag[index][i]
maxIndex = i
return maxIndex
return self.maxClass()
def dumpLabel(self,index:int,data:Instances):
text=""
text+=data.classAttribute().value(self.m_distribution.maxClass(index))
text+=" ("+str(Utils.roundDouble(self.m_distribution.perBag(index), 2))
if Utils.gr(self.m_distribution.numIncorrect(index), 0):
text+="/"+str(Utils.roundDouble(self.m_distribution.numIncorrect(index), 2))
text+=")"
return text
def handleNumericAttribute(self, trainInstances: Instances):
next = 1
last = 0
splitIndex = -1
self.m_distribution = Distribution(2, trainInstances.numClasses())
i = 0
for inst in trainInstances:
if inst.isMissing(self.m_attIndex):
break
self.m_distribution.add(1, inst)
i += 1
firstMiss = i
minSplit = 0.1 * self.m_distribution.total(
) / trainInstances.numClasses()
if Utils.gr(self.m_minNoObj, minSplit) or Utils.equal(
minSplit, self.m_minNoObj):
minSplit = self.m_minNoObj
elif Utils.gr(minSplit, 25):
minSplit = 25
if Utils.gr(2 * minSplit, firstMiss):
return
defaultEnt = self.infoGainCrit.oldEnt(self.m_distribution)
print("dfalut", defaultEnt)
while next < firstMiss:
if trainInstances.instance(next - 1).value(
self.m_attIndex) + 1e-5 < trainInstances.instance(
next).value(self.m_attIndex):
self.m_distribution.shiftRange(1, 0, trainInstances, last,
next)
if (Utils.gr(self.m_distribution.perBag(0), minSplit) or Utils.equal(self.m_distribution.perBag(0), minSplit))\
and (Utils.gr(self.m_distribution.perBag(1), minSplit) or Utils.equal(self.m_distribution.perBag(1), minSplit)):
currentInfoGain = self.infoGainCrit.splitCritValue(
self.m_distribution, self.m_sumOfWeights, defaultEnt)
if Utils.gr(currentInfoGain, self.m_infoGain):
self.m_infoGain = currentInfoGain
splitIndex = next - 1
self.m_index += 1
last = next
next += 1
if self.m_index == 0:
return
if self.m_useMDLcorrection:
self.m_infoGain = self.m_infoGain - (Utils.log2(self.m_index) /
self.m_sumOfWeights)
if Utils.gr(0, self.m_infoGain) or Utils.equal(0, self.m_infoGain):
return
self.m_numSubsets = 2
self.m_splitPoint = (
trainInstances.instance(splitIndex + 1).value(self.m_attIndex) +
trainInstances.instance(splitIndex).value(self.m_attIndex)) / 2
if self.m_splitPoint == trainInstances.instance(splitIndex + 1).value(
self.m_attIndex):
self.m_splitPoint = trainInstances.instance(splitIndex).value(
self.m_attIndex)
self.m_distribution = Distribution(2, trainInstances.numClasses())
self.m_distribution.addRange(0, trainInstances, 0, splitIndex + 1)
self.m_distribution.addRange(1, trainInstances, splitIndex + 1,
firstMiss)
self.m_gainRatio = self.gainRatioCrit.splitCritValue(
self.m_distribution, self.m_sumOfWeights, self.m_infoGain)
def maxBag(self):
max = 0
maxIndex = -1
for i in range(len(self.m_perBag)):
if Utils.gr(self.m_perBag[i], max) or Utils.equal(
self.m_perBag[i], max):
max = self.m_perBag[i]
maxIndex = i
return maxIndex
def check(self, minNoObj: float):
counter = 0
for i in range(len(self.m_perBag)):
if Utils.gr(self.m_perBag[i], minNoObj) or Utils.equal(
self.m_perBag[i], minNoObj):
counter += 1
if counter > 1:
return True
return False
def laplaceProb(self, classIndex: int, intIndex: int = None):
if intIndex is None:
return (self.m_perClass[classIndex] + 1) / (self.totaL +
len(self.m_perClass))
else:
if Utils.gr(self.m_perBag[intIndex], 0):
return (self.m_perClassPerBag[intIndex][classIndex] +
1) / (self.m_perBag[intIndex] + len(self.m_perClass))
return self.laplaceProb(classIndex)
def prune(self):
if not self.m_isLeaf:
for i in range(len(self.m_sons)):
self.son(i).prune()
if Utils.gr(self.errorsForTree(),
self.errorsForLeaf()) or Utils.equal(
self.errorsForTree(), self.errorsForLeaf()):
self.m_sons = None
self.m_isLeaf = None
self.m_localModel = NoSplit(self.localModel().distribution())
def prob(self, classIndex: int, intIndex: int = None):
if intIndex is None:
if not Utils.equal(self.totaL, 0):
return self.m_perClass[classIndex] / self.totaL
return 0
else:
if Utils.gr(self.m_perBag[intIndex], 0):
return self.m_perClassPerBag[intIndex][
classIndex] / self.m_perBag[intIndex]
return self.prob(classIndex)
def splitEnt(self, bags: Distribution, totalnoInst: float = None):
if totalnoInst is None:
return super().splitEnt(bags)
res = 0
noUnknown = totalnoInst - bags.total()
if Utils.gr(bags.total(), 0):
for i in range(bags.numBags()):
res = res - self.lnFunc(bags.perBag(i))
res = res - self.lnFunc(noUnknown)
res = res + self.lnFunc(totalnoInst)
return res / math.log(2)
def handleEnumeratedAttribute(self, instances: Instances):
numAttValues = instances.attribute(self.m_attIndex).numValues()
newDistribution = Distribution(numAttValues, instances.numClasses())
for inst in instances:
if not inst.isMissing(self.m_attIndex):
newDistribution.add(int(inst.value(self.m_attIndex)), inst)
self.m_distribution = newDistribution
for i in range(numAttValues):
if Utils.gr(newDistribution.perBag(i), self.m_minNoObj) or\
Utils.equal(newDistribution.perBag(i), self.m_minNoObj):
secondDistribution = Distribution(newDistribution, i)
if secondDistribution.check(self.m_minNoObj):
self.m_numSubsets = 2
currIG = self.infoGainCrit.splitCritValue(
secondDistribution, self.m_sumOfWeights)
currGR = self.gainRatioCrit.splitCritValue(
secondDistribution, self.m_sumOfWeights, currIG)
if i == 0 or Utils.gr(currGR, self.m_gainRatio):
self.m_gainRatio = currGR
self.m_infoGain = currIG
self.m_splitPoint = i
self.m_distribution = secondDistribution
def batchFinished(self):
if self.getInputFormat() is None:
raise Exception("No input instance format defined")
if self.m_ModesAndMeans is None:
sumOfWeights=self.getInputFormat().sumOfWeight()
counts=[[] for k in range(self.getInputFormat().numAttributes())]
for i in range(self.getInputFormat().numAttributes()):
if self.getInputFormat().attribute(i).isNominal():
counts[i]=[0]*self.getInputFormat().attribute(i).numValues()
if len(counts[i]) > 0:
counts[i][0]=sumOfWeights
sums=[]
for i in range(self.getInputFormat().numAttributes()):
sums.append(sumOfWeights)
results=[0]*self.getInputFormat().numAttributes()
for j in range(self.getInputFormat().numInstances()):
inst=self.getInputFormat().instance(j)
for i in range(inst.numValues()):
if not inst.isMissingSparse(i):
value=inst.valueSparse(i)
if inst.attributeSparse(i).isNominal():
if len(counts[inst.index(i)]) > 0:
counts[inst.index(i)][int(value)]+=inst.weight()
counts[inst.index(i)][0]-=inst.weight()
elif inst.attributeSparse(i).isNumeric():
results[inst.index(i)]+=inst.weight()*inst.valueSparse(i)
else:
if inst.attributeSparse(i).isNominal():
if len(counts[inst.index(i)]) > 0 :
counts[inst.index(i)][0]-=inst.weight()
elif inst.attributeSparse(i).isNumeric():
sums[inst.index(i)]-=inst.weight()
self.m_ModesAndMeans=[0]*self.getInputFormat().numAttributes()
for i in range(self.getInputFormat().numAttributes()):
if self.getInputFormat().attribute(i).isNominal():
if len(counts[i]) == 0:
self.m_ModesAndMeans[i]= Utils.missingValue()
else:
self.m_ModesAndMeans[i]= Utils.maxIndex(counts[i])
elif self.getInputFormat().attribute(i).isNumeric():
if Utils.gr(sums[i], 0):
self.m_ModesAndMeans[i]=results[i]/sums[i]
for i in range(self.getInputFormat().numInstances()):
self.convertInstance(self.getInputFormat().instance(i))
self.flushInput()
self.m_NewBatch=True
return self.numPendingOutput() != 0
def toSummaryString(self,printComplexityStatistics:bool,title:str="=== Summary ===\n"):
if printComplexityStatistics and self.m_NoPriors:
printComplexityStatistics=False
text=title+'\n'
if self.m_WithClass > 0:
if self.m_ClassIsNominal:
displayCorrect="correct" in self.m_metricsToDisplay
displayIncorrect="incorrect" in self.m_metricsToDisplay
displayKappa="kappa" in self.m_metricsToDisplay
if displayCorrect:
text+="Correctly Classified Instances "
text+= Utils.doubleToString(self.correct(), 12, 4) + " " + Utils.doubleToString(self.pctCorrect(), 12, 4) + " %\n"
if displayIncorrect:
text+="Incorrectly Classified Instances "
text+= Utils.doubleToString(self.incorrect(), 12, 4) + " " + Utils.doubleToString(self.pctIncorrect(), 12, 4) + " %\n"
if displayKappa:
text+="Kappa statistic "
text+= Utils.doubleToString(self.kappa(), 12, 4) + "\n"
if printComplexityStatistics:
displayKBRelative="kb relative" in self.m_metricsToDisplay
displayKBInfo="kb information" in self.m_metricsToDisplay
if displayKBRelative:
text+="K&B Relative Info Score "
text+= Utils.doubleToString(self.KBRelativeInformation(), 12, 4) + " %\n"
if displayKBInfo:
text+="K&B Information Score "
text+= Utils.doubleToString(self.KBInformation(), 12, 4) + " bits"
text+= Utils.doubleToString(self.KBMeanInformation(), 12, 4) + " bits/instance\n"
#if self.m_pluginMetrics != null:
else:
displayCorrelation="correlation" in self.m_metricsToDisplay
if displayCorrelation:
text+="Correlation coefficient "
text+= Utils.doubleToString(self.correlationCoefficient(), 12, 4) + "\n"
# if self.m_pluginMetrics != null:
if printComplexityStatistics and self.m_ComplexityStatisticsAvailable:
displayComplexityOrder0="complexity 0" in self.m_metricsToDisplay
displayComplexityScheme="complexity scheme" in self.m_metricsToDisplay
displayComplexityImprovement="complexity improvement" in self.m_metricsToDisplay
if displayComplexityOrder0:
text+="Class complexity | order 0 "
text+= Utils.doubleToString(self.SFPriorEntropy(), 12, 4) + " bits"
text+= Utils.doubleToString(self.SFMeanPriorEntropy(), 12, 4) + " bits/instance\n"
if displayComplexityScheme:
text+="Class complexity | scheme "
text+= Utils.doubleToString(self.SFSchemeEntropy(), 12, 4) + " bits"
text+= Utils.doubleToString(self.SFMeanSchemeEntropy(), 12, 4) + " bits/instance\n"
if displayComplexityImprovement:
text+="Complexity improvement (Sf) "
text+= Utils.doubleToString(self.SFEntropyGain(), 12, 4) + " bits"
text+= Utils.doubleToString(self.SFMeanEntropyGain(), 12, 4) + " bits/instance\n"
displayMAE = "mae" in self.m_metricsToDisplay
displayRMSE = "rmse" in self.m_metricsToDisplay
displayRAE = "rae" in self.m_metricsToDisplay
displayRRSE = "rrse" in self.m_metricsToDisplay
if displayMAE:
text+="Mean absolute error "
text+= Utils.doubleToString(self.meanAbsoluteError(), 12, 4) + "\n"
if displayRMSE:
text+="Root mean squared error "
text+= Utils.doubleToString(self.rootMeanSquaredError(), 12, 4) + "\n"
if not self.m_NoPriors:
if displayRAE:
text+="Relative absolute error "
text+= Utils.doubleToString(self.relativeAbsoluteError(), 12, 4) + " %\n"
if displayRRSE:
text+="Root relative squared error "
text+= Utils.doubleToString(self.rootRelativeSquaredError(), 12, 4) + " %\n"
if self.m_CoverageStatisticsAvailable:
displayCoverage="coverage" in self.m_metricsToDisplay
displayRegionSize="region size" in self.m_metricsToDisplay
if displayCoverage:
text+="Coverage of cases " + Utils.doubleToString(self.m_ConfLevel, 4, 2) + " level) "
text+= Utils.doubleToString(self.coverageOfTestCasesByPredictedRegions(), 12, 4) + " %\n"
if not self.m_NoPriors:
if displayRegionSize:
text+="Mean rel. region size (" + Utils.doubleToString(self.m_ConfLevel, 4, 2) + " level) "
text+= Utils.doubleToString(self.sizeOfPredictedRegions(), 12, 4) + " %\n"
if Utils.gr(self.unclassified(), 0):
text+="UnClassified Instances "
text+= Utils.doubleToString(self.unclassified(), 12, 4) + " " + Utils.doubleToString(self.pctUnclassified(), 12, 4) + " %\n"
text+="Total Number of Instances "
text+= Utils.doubleToString(self.m_WithClass, 12, 4) + "\n"
if self.m_MissingClass>0:
text+="Ignored Class Unknown Instances "
text+= Utils.doubleToString(self.m_MissingClass, 12, 4) + "\n"
return text