def decision_tree_classifier(self):
# define param_grid argument to give GridSearchCV
criterion = ['gini', 'entropy']
max_depth = range(1, 100, 2)
# get the best validated model
dtc = Decision_tree_classifier(x_train=self.x_train,
y_train=self.y_train,
cv=3,
n_jobs=6,
criterion=criterion,
max_depth=max_depth,
grid_search=True)
# print all possible parameter values and the best parameters
# dtc.print_parameter_candidates()
# dtc.print_best_estimator()
# return the accuracy score
return (dtc.evaluate(data=self.x_train,
targets=self.y_train,
average='macro'),
dtc.evaluate(data=self.x_test,
targets=self.y_test,
average='macro'))
def decision_tree_classifier(self):
max_depth = range(1, 14)
min_samples_leaf = range(1, 9)
dtc = Decision_tree_classifier(x_train=self.x_train,
y_train=self.y_train,
cv=3,
max_depth=max_depth,
min_samples_leaf=min_samples_leaf,
grid_search=True)
#dtc.print_parameter_candidates()
#dtc.print_best_estimator()
return (dtc.evaluate(data=self.x_train, targets=self.y_train),
dtc.evaluate(data=self.x_test, targets=self.y_test))
def decision_tree_classifier(self):
criterion = ['gini', 'entropy']
max_depth = range(1, 100, 2)
dtc = Decision_tree_classifier(x_train=self.x_train,
y_train=self.y_train,
cv=3,
n_jobs=6,
criterion=criterion,
max_depth=max_depth,
grid_search=True)
dtc.print_parameter_candidates()
dtc.print_best_estimator()
return (dtc.evaluate(data=self.x_train, targets=self.y_train),
dtc.evaluate(data=self.x_test, targets=self.y_test))
def decision_tree_classifier(self):
"""
for dtc, i train on the training data using different :
1) criterion
2) max_depth
:return: ((accuracy_train, recall_train, precision_train),
(accuracy_test, recall_test, precision_test))
"""
# define parameters
# criterion = ('gini', 'entropy')
# max_depth = np.logspace(start=1, stop=6, base=2, num=6, dtype=np.int)
# best result over all criterion: 'entropy'
# best result over all max_depth: 8
# scale down parameters around its best result
# criterion = ('gini', 'entropy')
# scale = 4
# max_depth = np.arange(start=8-scale, stop=8+scale, step=1, dtype=np.int)
# best result over all criterion: 'entropy'
# best result over all max_depth: 7
# Due to the reason that the tuned parameters's accuracy is much lower
# than the raw parameters, we would choose to use the raw parameters
# get the best validated model
dtc = Decision_tree_classifier(
x_train=self.x_train,
y_train=self.y_train,
# cv=5,
# criterion=criterion,
# max_depth=max_depth,
grid_search=True)
# print all possible parameter values and the best parameters
# dtc.print_parameter_candidates()
# dtc.print_best_estimator()
return (dtc.evaluate(data=self.x_train, targets=self.y_train),
dtc.evaluate(data=self.x_test, targets=self.y_test))
def decision_tree_classifier(self):
"""
for dtc, i train on the training data using different :
1) criterion
2) max_depth
:return: test accuracy of the dtc best model
"""
# define parameters
# criterion = ('gini', 'entropy')
# max_depth = np.logspace(start=1, stop=6, base=2, num=6, dtype=np.int)
# best result over all criterion: 'gini'
# best result over all max_depth: 4
# scale down parameters around its best result
criterion = ('gini', 'entropy')
scale = 1
max_depth = np.arange(start=4 - scale,
stop=4 + scale,
step=1,
dtype=np.int)
# best result over all criterion: 'gini'
# best result over all max_depth: 4
# get the best validated model
dtc = Decision_tree_classifier(x_train=self.x_train,
y_train=self.y_train,
cv=5,
criterion=criterion,
max_depth=max_depth,
grid_search=True)
# print all possible parameter values and the best parameters
# dtc.print_parameter_candidates()
# dtc.print_best_estimator()
# return the accuracy score
return (dtc.evaluate(data=self.x_train, targets=self.y_train),
dtc.evaluate(data=self.x_test, targets=self.y_test))