X = X.astype(float)
X = scale(X)
# remove useless features:
useless_features = (X.max(axis=0) - X.min(axis=0))==0
for j in reversed(range(n_features)):
if useless_features[j]:
X = np.delete(X, (j), axis=1)
X_train = X[:n_samples_train, :]
X_test = X[n_samples_train:, :]
y_train = y[:n_samples_train]
y_test = y[n_samples_train:]
### cross-val: ####
parameters = {'max_depth':[1000] , 'max_samples':[.1], 'max_features':[min(10, n_features)], 'n_estimators':[50]}
model = OneClassRF()
clf = grid_search.GridSearchCV(model, parameters, refit=False,cv=2)
clf.fit(X_train, y_train)
print('clf.best_params_', clf.best_params_)
model.set_params(**clf.best_params_)
print('OneClassRF processing...')
# weird: without CV but with same parameters, no error:
#model = OneClassRF(max_depth=1000, max_samples=.1, max_features=min(10, n_features), n_estimators=50, random_state=rng, n_jobs=-1) #commented since cross val
tstart = time()
### training only on normal data:
X_train = X_train[y_train==0]
y_train = y_train[y_train==0]
if useless_features[j]:
X = np.delete(X, (j), axis=1)
X_train = X[:n_samples_train, :]
X_test = X[n_samples_train:, :]
y_train = y[:n_samples_train]
y_test = y[n_samples_train:]
### cross-val: ####
parameters = {
'max_depth': [1000],
'max_samples': [.1],
'max_features': [min(10, n_features)],
'n_estimators': [50]
}
model = OneClassRF()
clf = grid_search.GridSearchCV(model, parameters, refit=False, cv=2)
clf.fit(X_train, y_train)
print('clf.best_params_', clf.best_params_)
model.set_params(**clf.best_params_)
print('OneClassRF processing...')
# weird: without CV but with same parameters, no error:
#model = OneClassRF(max_depth=1000, max_samples=.1, max_features=min(10, n_features), n_estimators=50, random_state=rng, n_jobs=-1) #commented since cross val
tstart = time()
### training only on normal data:
X_train = X_train[y_train == 0]
y_train = y_train[y_train == 0]
model.fit(X_train)
precision = np.zeros(n_axis)
fit_time = 0
predict_time = 0
try:
for ne in range(nb_exp):
print 'exp num:', ne
X, y = sh(X, y)
X_train = X[:n_samples_train, :]
X_test = X[n_samples_train:, :]
y_train = y[:n_samples_train]
y_test = y[n_samples_train:]
print('OneClassRF processing...')
model = OneClassRF()
# # training only on normal data: (not supported in cv)
# X_train = X_train[y_train == 0]
# y_train = y_train[y_train == 0]
tstart = time()
model.fit(X_train)
fit_time += time() - tstart
tstart = time()
scoring = model.predict(X_test) # the lower, the more normal
# scoring = scale(scoring)
predict_time += time() - tstart
fpr_, tpr_ = roc_curve(y_test, scoring)[:2]
f = interp1d(fpr_, tpr_)
import numpy as np
import matplotlib.pyplot as plt
from sklearn.ensemble import OneClassRF
from sklearn.preprocessing import scale
rng = np.random.RandomState(42)
# Generate train data
X = 0.3 * rng.randn(100, 2)
X_train = np.r_[X + 2, X - 2]
# fit the model
clf = OneClassRF(n_estimators=1, max_samples=1., max_depth=4, random_state=rng)
clf.fit(X_train)
# y_pred_train = -clf.decision_function(X_train)
# y_pred_test = -clf.decision_function(X_test)
# y_pred_outliers = -clf.decision_function(X_outliers)
# plot the line, the samples, and the nearest vectors to the plane
xx, yy = np.meshgrid(np.linspace(-5, 5, 200), np.linspace(-5, 5, 200))
Z = clf.decision_function(np.c_[xx.ravel(), yy.ravel()])
Z = Z.reshape(xx.shape)
#Z = scale(Z)
Z = np.log(Z)
#plt.title("OneClassRF")
levels = np.linspace(Z.min(), Z.max(), 1000)
plt.contourf(xx, yy, Z, cmap=plt.cm.Blues, levels=levels)
plt.scatter(X_train[:, 0], X_train[:, 1], c='white')
plt.axis('tight')
precision = np.zeros(n_axis)
fit_time = 0
predict_time = 0
try:
for ne in range(nb_exp):
print 'exp num:', ne
X, y = sh(X, y)
X_train = X[:n_samples_train, :]
X_test = X[n_samples_train:, :]
y_train = y[:n_samples_train]
y_test = y[n_samples_train:]
print('OneClassRF processing...')
model = OneClassRF()
# # training only on normal data: (not supported in cv)
# X_train = X_train[y_train == 0]
# y_train = y_train[y_train == 0]
tstart = time()
model.fit(X_train)
fit_time += time() - tstart
tstart = time()
scoring = model.predict(X_test) # the lower, the more normal
# scoring = scale(scoring)
predict_time += time() - tstart
fpr_, tpr_ = roc_curve(y_test, scoring)[:2]
f = interp1d(fpr_, tpr_)
