def train(self):
output = open(self.outputFilename,'w')
output_confusion = open("confusionMatrix.txt",'w')
accuracyList = []
precisionList = []
recallList = []
fMeasureList = []
if self.flag == "-r":
random = True
distance = False
diversity = False
probability = False
elif self.flag == "-d":
random = False
distance = True
diversity = False
probability = False
elif self.flag == "-v":
random = False
distance = False
diversity = True
probability = False
elif self.flag == "-p":
random = False
distance = False
diversity = False
probability = True
for i in range(0,5):#run 5 times to obtain an average
mySplitter = FileSplitter(self.inputFilename)
if self.inputFilename == "shuffled_magic04.data": #declare a classifier with best parameters for dataset
clf = svm.SVC(gamma=.01, C=100.0)
elif self.inputFilename == "ionosphereShuffle.txt":
clf = svm.SVC(gamma=0.10000000000000001, C=1.0)
elif self.inputFilename == "sensorShuffled.data":
clf = svm.SVC(gamma=.01, C=100.0)
elif self.inputFilename == "mushroom.data":
clf = svm.SVC(gamma=.10000000000000001, C=1.0)
trainContent = []
trainTracker = 0
accuracyIteration = []
precisionIteration = []
recallIteration = []
fMeasureIteration = []
for j in range(0,10):
trainTracker += 1
if j == 0:
trainContent.extend(mySplitter.removeInitialFive())
elif (j == 1) or (j == 2):
remainingContent = mySplitter.getContent()
if self.inputFilename == "shuffled_magic04.data":
x2, y2 = Formatter.separate_magic(remainingContent)
dec = clf.decision_function(x2)
elif self.inputFilename == "ionosphereShuffle.txt":
x2, y2 = Formatter.separate_ionosphere(remainingContent)
dec = clf.decision_function(x2)
elif self.inputFilename == "sensorShuffled.data":
x2, y2 = Formatter.separate_sensor(remainingContent)
dec = clf.decision_function(x2)
elif self.inputFilename == "mushroom.data":
x2, y2 = Formatter.separate_mushroom(remainingContent)
dec = clf.decision_function(x2)
if random == True:
trainContent.extend(mySplitter.removeRandom(.025))
elif distance == True:
trainContent.extend(mySplitter.removeClosest(dec,.025))
elif diversity == True:
trainContent.extend(mySplitter.removeBrinkers(dec,0.85,.025))
elif probability == True:
trainContent.extend(mySplitter.removeProbable(dec,.025))
else:
remainingContent = mySplitter.getContent()
if self.inputFilename == "shuffled_magic04.data":
x2, y2 = Formatter.separate_magic(remainingContent)
dec = clf.decision_function(x2)
elif self.inputFilename == "ionosphereShuffle.txt":
x2, y2 = Formatter.separate_ionosphere(remainingContent)
dec = clf.decision_function(x2)
elif self.inputFilename == "sensorShuffled.data":
x2, y2 = Formatter.separate_sensor(remainingContent)
dec = clf.decision_function(x2)
elif self.inputFilename == "mushroom.data":
x2, y2 = Formatter.separate_mushroom(remainingContent)
dec = clf.decision_function(x2)
if random == True:
trainContent.extend(mySplitter.removeRandom(.1))
elif distance == True:
trainContent.extend(mySplitter.removeClosest(dec,.1))
elif diversity == True:
trainContent.extend(mySplitter.removeBrinkers(dec,0.85,.1))
elif probability == True:
trainContent.extend(mySplitter.removeProbable(dec,.1))
if self.inputFilename == "shuffled_magic04.data":
x1,y1 = Formatter.separate_magic(trainContent) #attributes are stored in x, labels in y
predictContent = mySplitter.getTestContent()
x2,y2 = Formatter.separate_magic(predictContent)
elif self.inputFilename == "ionosphereShuffle.txt":
x1,y1 = Formatter.separate_ionosphere(trainContent)
predictContent = mySplitter.getTestContent()
x2,y2 = Formatter.separate_ionosphere(predictContent)
elif self.inputFilename == "sensorShuffled.data":
x1,y1 = Formatter.separate_sensor(trainContent)
predictContent = mySplitter.getTestContent()
x2,y2 = Formatter.separate_sensor(predictContent)
elif self.inputFilename == "mushroom.data":
x1,y1 = Formatter.separate_mushroom(trainContent)
predictContent = mySplitter.getTestContent()
x2,y2 = Formatter.separate_mushroom(predictContent)
clf.fit(x1,y1) #train the classifier with a labeled set
prediction = clf.predict(x2) #predict the labels for the x2 set, and store them in a variable
accuracy, precision, recall = Formatter.confusion_matrix(prediction, y2, output_confusion, j)
fMeasure = Formatter.f_measure(precision,recall)
accuracy *= 100
precision *= 100
recall *= 100
print ("Accuracy: " + str(accuracy) + "%")
print ("Precision: " + str(precision) + "%")
print ("Recall: " + str(recall) + "%")
print ("F-measure: " + str(fMeasure))
fMeasureIteration.append(fMeasure)
accuracyIteration.append(accuracy)
precisionIteration.append(precision)
recallIteration.append(recall)
fMeasureList.append(fMeasureIteration)
accuracyList.append(accuracyIteration)
precisionList.append(precisionIteration)
recallList.append(recallIteration)
outputList = Formatter.takeAverage(fMeasureList)
Formatter.writeHistogramOutput(outputList,output)
output.close()
output_confusion.close()
