def feature_bagging(attr_list, reviews_list, test_data, number_of_trees):
forest = []
for i in range(number_of_trees):
print(i)
rn.shuffle(attr_list)
var = tc.decision_tree(list(reviews_list.keys()), None, None)
mat = return_mat(attr_list, reviews_list)
prev = []
var.MakeTree(mat, attr_list, prev)
forest.append(var)
error = 0
for data in test_data:
actual_label = test_data[data][0]
prediction = None
pred = 0
for tree in forest:
p = predict(tree, test_data[data][1])
if p == 1:
pred = pred + 1
if (len(forest) - pred) > pred:
prediction = -1
else:
prediction = 1
if actual_label != prediction:
error = error + 1
print("error: ", error / len(test_data))
def feature_bagging(attr_list, reviews_list, test_data, train_set,
number_of_trees):
forest = []
for i in range(number_of_trees):
rn.shuffle(attr_list)
var = tc.decision_tree(list(reviews_list.keys()), None, None)
mat = return_mat(attr_list, reviews_list)
prev = []
var.MakeTree(mat, attr_list, prev)
forest.append(var)
print("Number Of tree: ", number_of_trees)
print("train error: ", return_dataset_error(reviews_list, forest))
print("test error: ", return_dataset_error(test_data, forest))
def DecisionTree(attr_list, reviews_list):
level_tree = {}
print("start making tree")
m = open('./model.pickle', 'wb')
mat = return_mat(attr_list, reviews_list)
var = tc.decision_tree(list(reviews_list.keys()), None, None)
start = timeit.default_timer()
prev = []
var.MakeTree(mat, attr_list, prev)
print(timeit.default_timer() - start)
pickle.dump(var, m)
m.close()
PrintTree(var)
CheckModel(var, "./train/train_dataset.txt")
CheckModel(var, "./test/test_dataset.txt")
def EarlyStoppingDT(mat, attr_list, reviews_list, depth, percentage_review,
ratio):
level_tree = {}
print("start making tree")
m = open('./model_earlystopping.pickle', 'wb')
var = tc.decision_tree(list(reviews_list.keys()), None, None)
prev = []
var.MakeTree(mat,
attr_list,
prev,
reviews_list=reviews_list,
depth=depth,
percentage_review=percentage_review,
ratio=ratio)
pickle.dump(var, m)
m.close()
PrintTree(var)
CheckModel(var, "./train/train_dataset.txt")
CheckModel(var, "./test/test_dataset.txt")
def noise_add(noise_percentage, attr_list, reviews_list):
total_noise = (noise_percentage / 100) * len(reviews_list)
while total_noise > 0:
r = rn.randint(0, len(reviews_list) - 1)
if reviews_list[r][0] >= 7:
reviews_list[r] = (-1, reviews_list[r][1])
else:
reviews_list[r] = (1, reviews_list[r][1])
total_noise = total_noise - 1
mat = return_mat(attr_list, reviews_list)
level_tree = {}
print("start making tree")
m = open('./noise_model.pickle', 'wb')
var = tc.decision_tree(list(reviews_list.keys()), None, None)
prev = []
var.MakeTree(mat, attr_list, prev)
pickle.dump(var, m)
m.close()
if var:
depth = 0
var.printTree(depth, level_tree)
for level in sorted(level_tree.keys()):
print(level, " : ", end=" ")
for value in level_tree[level]:
print(value.attr, "\t", value.label, end="\t")
print("\n")
CheckModel(var, "./train/train_dataset.txt")
CheckModel(var, "./test/test_dataset.txt")
