def testGenerate(self):
generate = DecisionTree.generate(5, 50)
learner = generate()
learner.learnModel(self.X, self.y)
self.assertEquals(learner.getMaxDepth(), 5)
self.assertEquals(learner.getMinSplit(), 50)
def testPredict(self):
decisionTree = DecisionTree()
decisionTree.learnModel(self.X, self.y)
predY = decisionTree.predict(self.X)
inds = numpy.random.permutation(self.X.shape[0])
predY2 = decisionTree.predict(self.X[inds, :])
self.assertTrue((predY[inds] == predY2).all())
#Let's test on -1, +1 labels
y2 = (self.y*2)-1
decisionTree.learnModel(self.X, y2)
predY2 = decisionTree.predict(self.X)
self.assertTrue((predY2 == predY*2-1).all())
def testParallelPenaltyGrid(self):
folds = 3
idx = Sampling.crossValidation(folds, self.X.shape[0])
decisionTree = DecisionTree()
bestLearner, meanErrors = decisionTree.parallelVfcv(self.X, self.y, idx)
trainX = self.X[0:40, :]
trainY = self.y[0:40]
paramDict = {}
paramDict["setMinSplit"] = decisionTree.getMinSplits()
paramDict["setMaxDepth"] = decisionTree.getMaxDepths()
idealPenalties = decisionTree.parallelPenaltyGrid(trainX, trainY, self.X, self.y, paramDict)
def testPredict2(self):
#We play around with parameters to maximise AUC on the IGF1_0-Haar data
dataDir = PathDefaults.getDataDir()
fileName = dataDir + "IGF1_0-Haar.npy"
XY = numpy.load(fileName)
X = XY[:, 0:XY.shape[1]-1]
y = XY[:, XY.shape[1]-1].ravel()
weight = numpy.bincount(numpy.array(y, numpy.int))[0]/float(y.shape[0])
#weight = 0.5
#weight = 0.9
folds = 3
decisionTree = DecisionTree()
decisionTree.setWeight(weight)
decisionTree.setMaxDepth(50)
#decisionTree.setMinSplit(100)
mean, var = decisionTree.evaluateCv(X, y, folds, Evaluator.auc)
logging.debug("AUC = " + str(mean))
logging.debug("Var = " + str(var))
def testSetWeight(self):
#Try weight = 0 and weight = 1
decisionTree = DecisionTree()
decisionTree.setWeight(0.0)
decisionTree.learnModel(self.X, self.y)
predY = decisionTree.predict(self.X)
self.assertTrue((predY == numpy.zeros(predY.shape[0])).all())
decisionTree.setWeight(1.0)
decisionTree.learnModel(self.X, self.y)
predY = decisionTree.predict(self.X)
self.assertTrue((predY == numpy.ones(predY.shape[0])).all())
def testMinSplit(self):
decisionTree = DecisionTree()
decisionTree.setMinSplit(20)
decisionTree.learnModel(self.X, self.y)
size = decisionTree.getTree().node_count
#orngTree.printTree(decisionTree.getClassifier())
decisionTree.setMinSplit(1)
decisionTree.learnModel(self.X, self.y)
size2 = decisionTree.getTree().node_count
#orngTree.printTree(decisionTree.getClassifier())
self.assertTrue(size < size2)
def testLearnModel(self):
decisionTree = DecisionTree()
decisionTree.learnModel(self.X, self.y)
tree = decisionTree.getClassifier()
def testParallelVfcv(self):
folds = 3
idx = Sampling.crossValidation(folds, self.X.shape[0])
decisionTree = DecisionTree()
bestLearner, meanErrors = decisionTree.parallelVfcv(self.X, self.y, idx)
def testSetMaxDepth(self):
maxDepth = 20
decisionTree = DecisionTree()
decisionTree.setMaxDepth(maxDepth)
decisionTree.learnModel(self.X, self.y)
#self.assertTrue(DecisionTree.depth(decisionTree.getClassifier().tree) <= maxDepth+1)
maxDepth = 5
decisionTree = DecisionTree()
decisionTree.setMaxDepth(maxDepth)
decisionTree.learnModel(self.X, self.y)
