def __init__(self, name):
self.model = None
if name == Constants.logistic:
self.model = LogisticRegression()
elif name == Constants.linear:
self.model = LinearRegression()
elif name == Constants.GBM:
self.model = GradientBoosting()
elif name == Constants.naiveBayes:
self.model = NaiveBayes()
elif name == Constants.decisionTree:
self.model = DecisionTree()
elif name == Constants.randomForest:
self.model = RandomForest()
elif name == Constants.SVM:
self.model = SVM()
elif name == Constants.xgboost:
self.model = XGBoost()
def trainClassifiers(data, algorithm):
x, t = Initializing.processData(data)
labels = np.unique(t)
listW = []
listB = []
for i in range(len(labels)):
tmpData = processDataSetPerClass(data, i, algorithm)
tmpX, tmpT = Initializing.processData(tmpData)
w, b = Initializing.initialParam(tmpX)
if algorithm == 'perceptron':
w, b = Perceptron.train(tmpX, tmpT, w, b)
elif algorithm == 'logistic':
# w, b = LogisticRegression.train(tmpX, tmpT, w, b)
LogisticRegression.bestW, LogisticRegression.bestB = Initializing.initialParam(tmpX)
trainingSet, testSet = CrossValidation.makeSets(tmpData) # Napravimo trening i test set
kTrainingSets, kValidSets = CrossValidation.kCrossValidationMakeSets(trainingSet, 5) # Napravimo k trening i test set-ova unakrsnom validacijom (k = 5)
w, b = LogisticRegression.crossTrain(kTrainingSets, kValidSets) # Istreniramo k trening setova i kao rezultat vratimo najbolje w i najbolje b (ono w i b za koje je greska bila najmanja)
else:
trainingSet, testSet = CrossValidation.makeSets(tmpData) # Napravimo trening i test set
kTrainingSets, kValidSets = CrossValidation.kCrossValidationMakeSets(trainingSet, 5) # Napravimo k trening i test set-ova unakrsnom validacijom (k = 5)
c = PassiveAggressiveAlgorithm.optC(kTrainingSets, kValidSets) # Podesimo optimalni parametar c
w, b = PassiveAggressiveAlgorithm.crossTrain(kTrainingSets, kValidSets, c) # Istreniramo k trening setova i kao rezultat vratimo najbolje w i najbolje b (ono w i b za koje je greska bila najmanja)
listW.append(np.array(w[0]).reshape(1,len(w[0])))
listB.append(np.array(b[0]).reshape(1,1))
return [listW, listB]
def plotLogisticRegression(self):
PlotInWindow.fig = Figure()
self.labelTitle["text"] = "Logistic Regression Algorithm"
file = self.labelForFile["text"]
file = ReadingFromFile.checkFile(file, "logistic")
fig = LogisticRegression.logisticRegressionPlotInWindow(file)
self.drawFigure(fig)
def plotAlgorithmInWindow(file, algorithm):
data = ReadingFromFile.readDataFromFile(file, ',') # Podaci ucitani iz fajla
trainingSet, testSet = CrossValidation.makeSets(data)
x, t = Initializing.processData(trainingSet)
t = Initializing.checkLabels(t, algorithm)
kTrainingSets, kValidSets = CrossValidation.kCrossValidationMakeSets(trainingSet, 5) # Napravimo k trening i test set-ova unakrsnom validacijom (k = 5)
if algorithm == "perceptron":
w, b = Initializing.initialParam(x)
w, b = Perceptron.train(x, t, w, b)
elif algorithm == "logistic":
w, b = LogisticRegression.crossTrain(kTrainingSets,kValidSets) # Istreniramo k trening setova i kao rezultat vratimo najbolje w i najbolje b (ono w i b za koje je greska bila najmanja)
else:
c = PassiveAggressiveAlgorithm.optC(kTrainingSets, kValidSets) # Podesimo optimalni parametar c
w, b = PassiveAggressiveAlgorithm.crossTrain(kTrainingSets, kValidSets, c) # Istreniramo k trening setova i kao rezultat vratimo najbolje w i najbolje b (ono w i b za koje je greska bila najmanja)
xTest, tTest = Initializing.processData(testSet)
fig = PlotInWindow.plotInWindow(xTest, w, b, tTest)
return fig
def logisticRegressionExample(file):
file = ReadingFromFile.checkFile(file, "logistic")
data = ReadingFromFile.readDataFromFile(file,
',') # Podaci ucitani iz fajla
trainingSet, testSet = CrossValidation.makeSets(
data) # Napravimo trening i test set
kTrainingSets, kValidSets = CrossValidation.kCrossValidationMakeSets(
trainingSet,
5) # Napravimo k trening i test set-ova unakrsnom validacijom (k = 5)
w, b = LogisticRegression.crossTrain(
kTrainingSets, kValidSets
) # Istreniramo k trening setova i kao rezultat vratimo najbolje w i najbolje b (ono w i b za koje je greska bila najmanja)
x, t = Initializing.processData(
testSet) # Rezultat crtamo i merimo nad test skupom podataka
t = Initializing.checkLabels(t, "logistic")
Plot.plotData(x, t)
Plot.plotLine(x, w, b)
plt.show()
# __Date__ : 1/5/2020. # __Author__ : CodePerfectPlus # __Package__ : Python 3 # __GitHub__ : https://www.github.com/codeperfectplus # from Algorithms import LogisticRegression lr = LogisticRegression() X = [2, 3, 5] y = [3, 5, 3] lr.fit(X, y)