def train(self, train_set, label, N_feat = 5, N_tree = 50, max_depth = 10):
""" Trains the forest, and stores trees for prediction
Arguments:
train_set - pandas.DataFrame containing the training data
label - Name of the label column in the DataFrame
N_feat - Number of features to consider for each trees
Default = 5
N_tree - Number of trees to use in the forest
Default = 50
max_depth - Maximum depth of trees to create
Default = 10
"""
self.trees = []
self.N_feat = N_feat
self.N_tree = N_tree
for i in range(N_tree):
t_set = train_set.copy()
t_set = reduce_features(self.N_feat, t_set, label)
self.trees.append(dt.get_tree(t_set, label, max_depth))
def LoadAndTreeEval():
""" Load the data and create a decision tree
Returns:
DTree generated for the data
"""
train, test, tune = l.CreateDataFrames('data/Build229Data.txt', 'data/FeatureNames.txt')
tree = dt.get_tree(train, 'Winner', 10)
p.dump(tree, open('classifiers/dtree03.p','wb'))
pred = tree.predict(test)
true = test['Winner'].values
print an.accuracy(pred, true)
print an.f1_score(pred, true)
return tree
def train(self, train_set, label, M = 10):
""" Trains the bagger and stores the resulting trees
Arguments:
train_set - pandas.DataFrame containing the training data
label - Name of the label column in the DataFrame
M - Number of trees to generate
Default = 10
"""
N = train_set.shape[0]
self.M = M
self.trees = []
for i in range(M):
idxs = an.bootstrap(range(N))
b_set = train_set.ix[idxs]
b_set.index = np.arange(N)
self.trees.append(dt.get_tree(b_set, label))
