def run_bagged_forest_prediction(examples_bt, test_examples_bt):
num_correct = 0
print("accuracy for bagged trees: training")
for i, example in enumerate(examples_bt):
print(i)
decision = 0
for dtree in dtrees:
decision += use_tree(dtree, example)
if decision > 0 and example.label == 1:
num_correct += 1
elif decision <= 0 and example.label == -1:
num_correct += 1
print(num_correct / len(examples_bt))
num_correct = 0
print("accuracy for bagged trees: test")
for example in test_examples_bt:
decision = 0
for dtree in dtrees:
decision += use_tree(dtree, example)
if decision > 0 and example.label == 1:
num_correct += 1
elif decision <= 0 and example.label == -1:
num_correct += 1
print(num_correct / len(test_examples_bt))
def transform_trees_to_feats(examples, trees): example_list = list(examples) new_list = [] for example in example_list: feat_array = np.array([use_tree(tree, example) for tree in trees]) new_list.append(LabeledEx(example.label, feat_array)) return new_list
def calc_error_target(stump, weights, examples, target_label):
error = 0
right_or_wrong = []
for i, ex in enumerate(examples):
guess = use_tree(stump, ex)
if guess != ex.label and guess != -1:
error += weights[i]
right_or_wrong.append(0)
elif guess == -1 and ex.label == target_label:
error += weights[i]
right_or_wrong.append(0)
else:
right_or_wrong.append(1)
return error, right_or_wrong
def use_classifier_target(H, examples, target_label):
num_correct = 0
if len(H) == 0:
return 0
for ex in examples:
sign = 0
for (alpha, h) in H:
guess = use_tree(h, ex)
if guess == target_label:
sign += alpha
else:
sign -= alpha
if sign >= 0 and ex.label == target_label:
num_correct += 1
elif sign < 0 and ex.label != target_label:
num_correct += 1
return num_correct
def use_classifier_multiclass(H, examples, labels):
num_correct = 0
for ex in examples:
best = [-1000, None]
for label in labels:
sign = 0
for (alpha, h, h_label) in H:
guess = use_tree(h, ex)
if guess == h_label:
sign += alpha
else:
sign -= alpha
if sign > best[0]:
best[0] = sign
best[1] = label
if best[1] == ex.label:
num_correct += 1
return num_correct
def transform_trees_to_feats(examples, trees):
example_list = list(examples)
for example in example_list:
example.feats = np.array([use_tree(tree, example) for tree in trees])
return example_list