proxies = {
'http': args.http_proxy,
'https': args.https_proxy
}
print(proxies)
# binary classification: Beta prior
prior = (1, 1)
# Bayesian decision tree parameters
partition_prior = 0.9
delta = 0
root = BinaryClassificationNode(partition_prior, prior)
# training/test data
train, test = load_ripley(proxies)
# train
X_train = train[:, :-1]
y_train = train[:, -1]
X_test = test[:, :-1]
y_test = test[:, -1]
root.fit(X_train, y_train, delta)
print(root)
print()
print('Tree depth and number of leaves:', root.depth_and_leaves())
# compute accuracy
prediction_train = root.predict(X_train)
prediction_test = root.predict(X_test)
accuracy_train = (prediction_train == y_train).mean()
# y_train = np.zeros((n, 1))
# y_train[(X_train[:, 0] >= 1) & (X_train[:, 0] < 2) & (X_train[:, 1] <= 3)] = 1
# y_train[(X_train[:, 0] >= 2) & (X_train[:, 0] < 3) & (X_train[:, 1] <= 1)] = 1
# y_train[(X_train[:, 0] >= 3)] = 1
#
# angle = 30*np.pi/180
# X_train_rot = X_train.copy()
# X_train_rot[:, 0] = np.cos(angle)*X_train[:, 0] + np.sin(angle)*X_train[:, 1]
# X_train_rot[:, 1] = -np.sin(angle)*X_train[:, 0] + np.cos(angle)*X_train[:, 1]
# X_train = X_train_rot
#
# train = np.hstack((X_train, y_train))
# test = train
# or, alternatively, load a UCI dataset
train, test = helper.load_ripley(proxies)
n_dim = len(np.unique(train[:, -1]))
if train is test:
# perform a 50:50 train:test split if no test data is given
train = train[0::2]
test = test[1::2]
X_train = train[:, :-1]
y_train = train[:, -1]
X_test = test[:, :-1]
y_test = test[:, -1]
# prior
prior_pseudo_observations = 1