def train_and_predict():
filename = '../data/6.5w_distinct1_labeled.csv'
data = pandas.read_csv(filename)
user_agent = n_gram_v2(filename, 2)
class_features = text_to_number(filename)
mouse_features = data_normalization(filename)
#mouse_features = data[['mouse_x', 'mouse_y']]
target = data.label
# 矩阵降维（n, 1）--> (n,)
target = target.ravel()
all_features = numpy.concatenate(
(class_features, mouse_features, user_agent), axis=1)
# 样本随机
all_features, target = shuffle(all_features, target, random_state=0)
x_train, x_test, y_train, y_test = train_test_split(all_features,
target,
test_size=0.4,
random_state=0)
classifier = RandomForestClassifier(n_jobs=-1,
max_depth=47,
min_samples_split=2,
min_samples_leaf=1,
max_features=31,
n_estimators=180)
'''
classifier.fit(x_train, y_train)
joblib.dump(classifier, '../models/RandomForest.m')
y_predict = classifier.predict(x_test)
y_predict_proba = classifier.predict_proba(x_test)
do_metrics(y_test, y_predict, y_predict_proba[:, 1])
'''
scores = cross_val_score(classifier, all_features, target, cv=10)
print(scores)
print(numpy.mean(scores))
def cross_validate():
filename = '../data/6.5w_distinct1_labeled.csv'
data = pandas.read_csv(filename)
user_agent = n_gram_v2(filename, 3)
class_features = text_to_number(filename)
mouse_features = data_normalization(filename)
target = data.label
# 矩阵降维（n, 1）--> (n,)
target = target.ravel()
all_features = numpy.concatenate((class_features, mouse_features), axis=1)
# 样本随机
all_features, target = shuffle(all_features, target, random_state=0)
#x_train, x_test, y_train, y_test = train_test_split(all_features, target, test_size = 0.4, random_state = 0)
classifier = LogisticRegression()
scores = cross_val_score(classifier, all_features, target, cv=10)
print(scores)
print(numpy.mean(scores))
def train_and_predict():
filename = '../data/6.5w_distinct1_labeled.csv'
data = pandas.read_csv(filename)
user_agent = n_gram_v2(filename, 1)
class_features = text_to_number(filename)
mouse_features = data_normalization(filename)
#mouse_features = data[['mouse_x', 'mouse_y']]
target = data.label
# 矩阵降维（n, 1）--> (n,)
target = target.ravel()
all_features = numpy.concatenate((class_features, mouse_features, user_agent), axis = 1)
# 样本随机
all_features, target = shuffle(all_features, target, random_state = 0)
x_train, x_test, y_train, y_test = train_test_split(all_features, target, test_size = 0.4, random_state = 0)
classifier = KNeighborsClassifier(n_neighbors = 200)
classifier.fit(x_train, y_train)
joblib.dump(classifier, '../models/KNN.m')
y_predict = classifier.predict(x_test)
y_predict_proba = classifier.predict_proba(x_test)
do_metrics(y_test, y_predict, y_predict_proba[:, 1])
def write_features_to_file():
filename = '../data/6.5w_distinct1_labeled.csv'
data = pandas.read_csv(filename)
user_agent = n_gram_v2(filename, 2)
# 所有特征放入DataFrame
all_features = user_agent
class_features = text_to_number(filename)
class_features_names = [
'event_action_account_input', 'event_action_pwd_input',
'event_action_login_click', 'java_enable', 'java_disable',
'cookie_enable', 'cookie_disenable'
]
count = 0
for name in class_features_names:
all_features[name] = class_features[:, count]
count += 1
#mouse_features = data[['mouse_x', 'mouse_y']]
mouse_features = data_normalization(filename)
all_features['mouse_x'] = mouse_features[:, 0]
all_features['mouse_y'] = mouse_features[:, 1]
all_features.to_csv('../data/6.5w_distinct1_features.csv')
def random_forest():
filename = '../data/6.5w_distinct1_labeled.csv'
data = pandas.read_csv(filename)
user_agent = n_gram_v2(filename, 1)
# 所有特征放入DataFrame
all_features = user_agent
class_features = text_to_number(filename)
class_features_names = [
'event_action_account_input', 'event_action_pwd_input',
'event_action_login_click', 'java_enable', 'java_disable',
'cookie_enable', 'cookie_disenable'
]
count = 0
for name in class_features_names:
all_features[name] = class_features[:, count]
count += 1
#mouse_features = data[['mouse_x', 'mouse_y']]
mouse_features = data_normalization(filename)
all_features['mouse_x'] = mouse_features[:, 0]
all_features['mouse_y'] = mouse_features[:, 1]
#all_features['mouse_x'] = data['mouse_x']
#all_features['mouse_y'] = data['mouse_y']
target = data.label
# 矩阵降维（n, 1）--> (n,)
target = target.ravel()
#all_features = numpy.concatenate((class_features, mouse_features, user_agent), axis = 1)
# 样本随机
#all_features, target = shuffle(all_features, target, random_state = 0)
classifier = RandomForestClassifier()
classifier.fit(all_features, target)
feature_importances_indices = numpy.argsort(
classifier.feature_importances_)
for num in range(numpy.shape(all_features)[1]):
print('%d. %s %f' % (
num, all_features.columns.values[feature_importances_indices[num]],
classifier.feature_importances_[feature_importances_indices[num]]))