def main():
'''main method'''
targets = argv[argv.index("-target") + 1][1:-1].split(
',') #read targets from input
regression, advice = False, False
if "-reg" in argv:
regression = True
if "-expAdvice" in argv:
advice = True
for target in targets:
data = Utils.readTrainingData(target, regression,
advice) #read training data
numberOfTrees = 10 #number of trees for boosting
trees = [] #initialize place holder for trees
for i in range(numberOfTrees): #learn each tree and update gradient
print('=' * 20, "learning tree", str(i), '=' * 20)
node.setMaxDepth(2)
node.learnTree(data) #learn RRT
trees.append(node.learnedDecisionTree)
Boosting.updateGradients(data, trees)
for tree in trees:
print('=' * 30, "tree", str(trees.index(tree)), '=' * 30)
for clause in tree:
print(clause)
testData = Utils.readTestData(target, regression) #read testing data
Boosting.performInference(testData,
trees) #get probability of test examples
def learn_clf(self, facts, pos, neg, bk):
'''learns classifier'''
self.trees = {}
for target in self.targets:
data = Utils.setTrainingData(target=target,
facts=facts,
pos=pos,
neg=neg,
bk=bk,
regression=self.regression,
sampling_rate=self.sampling_rate)
trees = []
for i in range(self.numberOfTrees):
print('=' * 20, "learning tree", str(i + 1), '=' * 20)
node.setMaxDepth(self.treeDepth)
node.learnTree(data)
trees.append(node.learnedDecisionTree)
Boosting.updateGradients(data, trees)
self.trees[target] = trees
for target in self.targets:
if target in self.trees:
number_of_trees = len(self.trees[target])
counter = 1
for tree in self.trees[target]:
print('=' * 30, "tree", str(counter), '=' * 30)
for clause in tree:
print(clause)
counter += 1
def learn(self, facts, examples, bk):
"""Learn the clause."""
for target in self.targets:
data = Utils.setTrainingData(target=target,
facts=facts,
examples=examples,
bk=bk,
regression=self.regression,
sampling_rate=self.sampling_rate)
trees = []
treeLCAs = []
for i in range(self.numberOfTrees):
print('=' * 20, "learning tree", str(i), '=' * 20)
node.setMaxDepth(self.treeDepth)
node.learnTree(data)
trees.append(node.learnedDecisionTree)
treeLCAs.append(node.LCAS)
Boosting.updateGradients(data, trees, loss=self.loss)
self.trees[target] = trees
self.treeLCAs[target] = treeLCAs
for tree in trees:
print('=' * 30, "tree", str(trees.index(tree)), '=' * 30)
for clause in tree:
print(clause)
'''
def learn(self, facts=None, examples=None, bk=None, pos=None, neg=None):
'''learns the regression tree per target
before learning, make sure that the training and test
and bk files are all in the default format.
change all formats as required as required during runtime.
'''
for target in self.targets: #for every target learn
data = Utils.setTrainingData(target=target,
facts=facts,
examples=examples,
bk=bk,
regression=self.regression,
sampling_rate=self.sampling_rate,
pos=pos,
neg=neg)
trees = [] #initialize place holder for trees
for i in range(self.numberOfTrees):
print('=' * 20, "learning tree", str(i), '=' * 20)
node.setMaxDepth(
self.treeDepth) #set max depth of individual tree learned
node.learnTree(data) #learn the regression tree
trees.append(node.learnedDecisionTree)
Boosting.updateGradients(data, trees, loss=self.loss)
self.trees[target] = trees
for tree in trees: #print each tree learned
print('=' * 30, "tree", str(trees.index(tree)), '=' * 30)
for clause in tree:
print(clause)
def infer(self,facts,examples):
self.testExamples = {}
for target in self.targets:
data = Utils.setTestData(target=target,facts=facts,examples=examples,regression=self.regression)
Boosting.performInference(data,self.trees[target])
self.testExamples[target] = data.examples
valueset = set(list(data.examples.values()))
print (data.examples)
def infer(self, facts, examples):
'''perform inference on learned trees'''
self.testExamples = {}
for target in self.targets:
data = Utils.setTestData(target=target,
facts=facts,
examples=examples,
regression=self.regression)
Boosting.performInference(data, self.trees[target])
self.testExamples[target] = data.examples
print(data.examples)
def main():
'''main method'''
parameters = Arguments().args
#print(parameters)
for target in parameters.target:
# Read the training data.
data = Utils.readTrainingData(target,
path=parameters.train,
regression=parameters.reg,
advice=parameters.expAdvice)
# Initialize an empty place holder for the trees.
trees = []
# Learn each tree and update the gradient.
for i in range(parameters.trees):
if parameters.verbose:
print('=' * 20, "learning tree", str(i), '=' * 20)
node.setMaxDepth(2)
node.learnTree(data) # Learn relational regression tree
trees.append(node.learnedDecisionTree)
Boosting.updateGradients(data, trees)
for tree in trees:
if parameters.verbose:
print('=' * 30, "tree", str(trees.index(tree)), '=' * 30)
for clause in tree:
print(clause)
# Read the testing data.
testData = Utils.readTestData(target,
path=parameters.test,
regression=parameters.reg)
# Get the probability of the test examples.
Boosting.performInference(testData, trees)
if parameters.reg:
# View test example values (for regression)
print(testData.examples)
else:
# View test query probabilities (for classification)
print(testData.pos)
print(testData.neg)
def grad_boost():
classification_data, classification_labels = datasets.make_classification(
n_samples=1000,
n_features=2,
n_informative=2,
n_classes=2,
n_redundant=0,
n_clusters_per_class=1,
random_state=23)
train_data, test_data, train_labels, test_labels = model_selection.train_test_split(
classification_data,
classification_labels,
test_size=0.3,
random_state=1)
n_trees = 10
max_depth = 60
num_leaves = 5
min_samples_leaf = 5
criteria = 'gini'
model_b = Boosting(n_trees=n_trees,
max_depth=max_depth,
num_leaves=num_leaves,
min_samples_leaf=min_samples_leaf,
criteria=criteria)
model_b.fit(train_data, train_labels)
train_answers = model_b.predict(train_data)
test_answers = model_b.predict(test_data)
# Точность на обучающей выборке
train_accuracy = accuracy_metric(train_labels, train_answers)
print('\n\nГрадиентный бустинг\n')
print(f'n_trees={n_trees}, max_depth={max_depth}, \
num_leaves={num_leaves}, \
min_samples_leaf={min_samples_leaf}, \
criteria={criteria}')
print(f' Точность на обучающей выборке: {train_accuracy:.3f}')
# Точность на тестовой выборке
test_accuracy = accuracy_metric(test_labels, test_answers)
print(f' Точность на тестовой выборке: {test_accuracy:.3f}')
if __name__ == '__main__':
print(
"Please select the number against the learner you want to build the model for datasets:"
)
print("1 -> Decision Tree")
print("2 -> K-NN")
print("3 -> Neural Networks")
print("4 -> Boosting")
print("5 -> SVM")
switcher = {
1: DecisionTree(),
2: KNN(),
3: NeuralNetworks(),
4: Boosting(),
5: SVM(),
}
datasets = dataset_loader.load_datasets()
model_number = input("Enter your value: ")
model = switcher.get(int(model_number), "Invalid learner")
print("Please select the dataset:")
print("1 -> Diabetes Retinatherapy")
print("2 -> Phishing Website")
dataset_number = input("Enter your value: ")
if int(dataset_number) == 1:
item = datasets[0]
else:
item = datasets[1]