def calculate(A, X, Y):
A = sp.coo_matrix(A)
A = A + A.T.multiply(A.T > A) - A.multiply(A.T > A)
rowsum = np.array(A.sum(1)).clip(min=1)
r_inv_sqrt = np.power(rowsum, -0.5).flatten()
r_mat_inv_sqrt = sp.diags(r_inv_sqrt)
A = A.dot(r_mat_inv_sqrt).transpose().dot(r_mat_inv_sqrt)
low = 0.5 * sp.eye(A.shape[0]) + A
high = 0.5 * sp.eye(A.shape[0]) - A
low = low.todense()
high = high.todense()
low_signal = np.dot(np.dot(low, low), X)
high_signal = np.dot(np.dot(high, high), X)
low_MLP = MLPClassifier(hidden_layer_sizes=(16),
activation='relu',
max_iter=2000)
low_MLP.fit(low_signal[:100, :], Y[:100])
low_pred = low_MLP.predict(low_signal[100:, :])
high_MLP = MLPClassifier(hidden_layer_sizes=(16),
activation='relu',
max_iter=2000)
high_MLP.fit(high_signal[:100, :], Y[:100])
high_pred = high_MLP.predict(high_signal[100:, :])
return acc(Y[100:], low_pred), acc(Y[100:], high_pred)
def on_epoch_end(self,epoch,logs=None):
x_1 = self.validation_data[0]
x_2 = self.validation_data[1]
y_test = self.validation_data[2]
print('Dims Validation data: %s %s %s',x_1.shape,x_2.shape,y_test.shape)
# predicting outputs for val data
y_pred = self.model.predict([x_1,x_2])
# selecting the top value of predictions and
y_test = np.argmax(y_test, axis=-1)
y_pred = np.argmax(y_pred, axis=-1)
self.acc_val.append(acc(y_test,y_pred))
print ('Acc: ',acc(y_test,y_pred))
def decision_tree_accuracy(X, y, random, depth, test, crit):
X_train, X_test, y_train, y_test = train_test_split(X,
y,
test_size=test,
random_state=random)
regressor = tree.DecisionTreeClassifier(criterion=crit,
max_depth=depth,
random_state=random)
regressor.fit(X_train, y_train)
ytr_pred = regressor.predict(X_train)
yts_pred = regressor.predict(X_test)
acc_ytr = acc(y_train, ytr_pred)
acc_yts = acc(y_test, yts_pred)
print(f'Accuracy test = {acc_yts}')
print(f'Accuracy train = {acc_ytr}')
def main():
x_train, y_train, x_test, y_test = get_data()
for n in [2, 3, 5, 10, 16]:
sfs = SFS(KNeighborsClassifier(n_neighbors=7),
k_features=n,
forward=False,
floating=True,
scoring='accuracy',
cv=0)
sfs = sfs.fit(x_train, y_train)
print('\nSequential Floating Forward Selection: ', n)
feat_cols = list(sfs.k_feature_idx_)
print(feat_cols)
knn = KNeighborsClassifier(n_neighbors=3)
knn.fit(x_train[:, feat_cols], y_train)
y_train_pred = knn.predict(x_train[:, feat_cols])
print('Training accuracy on selected features: %.3f' %
acc(y_train, y_train_pred))
y_test_pred = knn.predict(x_test[:, feat_cols])
print('Testing accuracy on selected features: %.3f' %
acc(y_test, y_test_pred))
print(confusion_matrix(y_test, y_test_pred))
print(classification_report(y_test, y_test_pred))
if n == 2:
fig, axs = plt.subplots(2)
fig.suptitle("SFS(KNN) Scatter Plot", fontsize='small')
axs[0].scatter(x_train[:, feat_cols[0]],
x_train[:, feat_cols[1]],
marker='o',
c=y_train,
s=25,
edgecolor='k')
axs[1].scatter(x_test[:, feat_cols[0]],
x_test[:, feat_cols[1]],
marker='o',
c=y_test,
s=25,
edgecolor='k')
plt.show()
def evaluate(model,
iterator_function,
_batch_count,
cuda_device,
output_buffer=sys.stderr):
if output_buffer is not None:
print(_batch_count, file=output_buffer)
model.eval()
with torch.no_grad():
predictions = []
expectations = []
batch_generator = range(_batch_count)
if output_buffer is not None:
batch_generator = tqdm(batch_generator)
for _ in batch_generator:
features, targets = iterator_function()
if cuda_device != -1:
features = features.cuda(device=cuda_device)
probs, _, _ = model(example_batch=features)
batch_pred = np.argmax(probs.detach().cpu().numpy(),
axis=-1).tolist()
batch_tgt = targets.detach().cpu().numpy().tolist()
predictions.extend(batch_pred)
expectations.extend(batch_tgt)
model.train()
return acc(expectations, predictions) * 100, \
pr(expectations, predictions) * 100, \
rc(expectations, predictions) * 100, \
f1(expectations, predictions) * 100,
def combinationPredict(predict, samples_test):
labels_samples, merger_min, merger_max, merger_sum, merger_pro = add.fusoesDiego(
predict, samples_test)
classSeg = np_utils.categorical_probas_to_classes(predict)
classMin = np_utils.categorical_probas_to_classes(merger_min)
classMax = np_utils.categorical_probas_to_classes(merger_max)
classSom = np_utils.categorical_probas_to_classes(merger_sum)
classPro = np_utils.categorical_probas_to_classes(merger_pro)
print()
print("Min: " + str(acc(labels_samples, classMin)))
print("Max: " + str(acc(labels_samples, classMax)))
print("Sum: " + str(acc(labels_samples, classSom)))
print("Product: " + str(acc(labels_samples, classPro)))
print()
def test_converter(self):
interpreter = tf.lite.Interpreter(
os.path.join(self.model_path, 'model.tflite'))
interpreter.allocate_tensors()
input_details = interpreter.get_input_details()
output_details = interpreter.get_output_details()
#load test data
[test_data, test_labels] = read_database(self.data_path)
test_data = test_data / 255.0
test_data = test_data[..., tf.newaxis].astype("float32")
predData = np.ndarray(shape=(test_data.shape[0]), dtype='uint8')
for i in range(0, test_data.shape[0]):
test_data_temp = np.array(test_data[[i], :, :, :], dtype='float32')
interpreter.set_tensor(input_details[0]['index'], test_data_temp)
interpreter.invoke()
output_data = interpreter.get_tensor(output_details[0]['index'])
predData[i] = np.argmax(output_data)
kappa = metrics.cohen_kappa_score(test_labels, predData)
print('Kappa:', kappa)
accuracy = acc(test_labels, predData)
print('Accuracy:', accuracy)
print('Confusion matrix:')
confusion_mat = metrics.confusion_matrix(test_labels, predData)
print(confusion_mat)
return (accuracy)
def actual_prediction_accuracy(self):
if self.predicted:
print("Prediction accuracy:", acc(self.y_test, self.y_pred))
else:
raise TypeError(
"Predictions for model are not available, use 'predict' method first!"
)
def predict(X_spam, X_ham, X_test, y_test):
pred = []
for X in X_test:
spam = nb(X_spam, X)
ham = nb(X_ham, X)
pred.append(1) if spam > ham else pred.append(0)
print(pred, y_test)
print('Accuracy:', acc(pred, y_test))
def cluster_acc(Y, clusterLabels):
assert (Y.shape == clusterLabels.shape)
pred = np.empty_like(Y)
for label in set(clusterLabels):
mask = clusterLabels == label
sub = Y[mask]
target = Counter(sub).most_common(1)[0][0]
pred[mask] = target
return acc(Y, pred)
def build_classifier_and_test(train_X,
train_y,
test_X,
test_y,
clf,
print_train_result=True):
clf.fit(train_X, train_y)
if print_train_result == True:
p_tr = clf.predict(train_X)
print("Train Accuracy:\t", acc(train_y, p_tr))
print("Train Precision:\t", pr(train_y, p_tr))
print("Train Recall_score:\t", rc(train_y, p_tr))
print("Train F-score:\t", f1(train_y, p_tr))
predicted = clf.predict(test_X)
print("Accuracy:\t", acc(test_y, predicted))
print("Precision:\t", pr(test_y, predicted))
print("Recall_score:\t", rc(test_y, predicted))
print("F-score:\t", f1(test_y, predicted))
def clone_analysis(data_paths):
code = []
labels = []
positives = 0
for file_name in data_paths:
data = json.load(open(file_name))
for example in data:
code.append(example['tokenized'])
l = 0
if 'label' in example.keys():
l = int(example['label'])
elif 'lebel' in example.keys():
l = int(example['lebel'])
elif 'leble' in example.keys():
l = int(example['leble'])
elif 'lable' in example.keys():
l = int(example['lable'])
if l > 1:
l = 1
positives += l
labels.append(l)
print(len(code), len(labels), positives, len(labels) - positives)
vectorizer = TfidfVectorizer(input=code,
lowercase=False,
ngram_range=(1, 3))
X = vectorizer.fit_transform(code)
model = KMeans(n_clusters=10, max_iter=100)
model.fit(X)
y = model.predict(X)
cluster_to_positive = [0] * 10
cluster_to_negative = [0] * 10
for pred, label in zip(y, labels):
if label == 1:
cluster_to_positive[pred] += 1
else:
cluster_to_negative[pred] += 1
print(cluster_to_positive)
print(cluster_to_negative)
percentages = [
float(p) / (p + n)
for p, n in zip(cluster_to_positive, cluster_to_negative)
]
for p in percentages:
print(p)
for _ in range(5):
XTrain, XTest, YTrain, YTest = train_test_split(X,
labels,
test_size=0.2)
model = RandomForestClassifier()
model.fit(XTrain, YTrain)
predicted = model.predict(XTest)
print('%.3f\t%.3f\t%.3f\t%.3f' %
(acc(YTest, predicted) * 100, pr(YTest, predicted) * 100,
rc(YTest, predicted) * 100, f1(YTest, predicted) * 100))
pass
def cluster_acc(Y, clusterLabels): #used in clustering.py
assert (Y.shape == clusterLabels.shape)
pred = np.empty_like(Y)
for label in set(clusterLabels):
mask = clusterLabels == label
sub = Y[mask]
target = Counter(sub).most_common(1)[0][0]
pred[mask] = target
# assert max(pred) == max(Y)
# assert min(pred) == min(Y)
return acc(Y, pred)
def cluster_acc(y, cluster_labels):
assert (y.shape == cluster_labels.shape)
pred = np.empty_like(y)
for label in set(cluster_labels):
mask = cluster_labels == label
sub = y[mask]
target = Counter(sub).most_common(1)[0][0]
pred[mask] = target
# assert max(pred) == max(Y)
# assert min(pred) == min(Y)
return acc(y, pred)
def cluster_acc(Y,clusterLabels):
assert (Y.shape == clusterLabels.shape)
pred = np.empty_like(Y)
for label in set(clusterLabels):
mask = clusterLabels == label
sub = Y[mask]
target = Counter(sub).most_common(1)[0][0]
pred[mask] = target
# assert max(pred) == max(Y)
# assert min(pred) == min(Y)
return acc(Y,pred)
def train(self):
x, y = np.load('images/64px_image_x.npy'), np.load(
'images/64px_image_y.npy')
x = np.reshape(x, (40000, 64, 64, 1))
kmeans = KMeans(n_clusters=2, n_init=20)
y_pred = kmeans.fit_predict(self.encoder.predict(x))
y_pred_last = np.copy(y_pred)
self.model.get_layer(name='clustering').set_weights(
[kmeans.cluster_centers_])
loss = 0
ae_loss = 0
index = 0
maxiter = 80000
update_interval = 100
index_array = np.arange(x.shape[0])
batch_size = 16
tol = 0.001
# model.load_weights('DEC_model_final.h5')
for ite in range(int(maxiter)):
if ite % update_interval == 0:
q = self.model.predict(x, verbose=0)
# update the auxiliary target distribution p
p = self.target_distribution(q)
# evaluate the clustering performance
y_pred = q.argmax(1)
if y is not None:
acc = np.round(metrics.acc(y, y_pred), 5)
nmi = np.round(metrics.nmi(y, y_pred), 5)
ari = np.round(metrics.ari(y, y_pred), 5)
loss = np.round(loss, 5)
print(
'Iter %d: acc = %.5f, nmi = %.5f, ari = %.5f, loss=%.5f'
% (ite, acc, nmi, ari, loss))
# check stop criterion - model convergence
delta_label = np.sum(y_pred != y_pred_last).astype(
np.float32) / y_pred.shape[0]
y_pred_last = np.copy(y_pred)
if ite > 0 and delta_label < tol:
print('delta_label ', delta_label, '< tol ', tol)
print('Reached tolerance threshold. Stopping training.')
break
idx = np.random.randint(low=0, high=x.shape[0], size=batch_size)
# ae_loss = ae.train_on_batch(x=x[idx], y=x[idx])
loss = self.model.train_on_batch(x=x[idx], y=p[idx])
index = index + 1 if (index + 1) * batch_size <= x.shape[0] else 0
self.model.save_weights('DEC_model_final_64px.h5')
self.test_model()
def on_epoch_end(self, epoch, logs=None):
if int(epochs/10) != 0 and epoch % int(epochs/10) != 0:
return
feature_model = Model(self.model.input,
self.model.get_layer(
'encoder_%d' % (int(len(self.model.layers) / 2) - 1)).output)
features = feature_model.predict(self.x)
km = KMeans(n_clusters=len(np.unique(self.y)), n_init=20, n_jobs=4)
y_pred = km.fit_predict(features)
# print()
print(' '*8 + '|==> acc: %.4f, nmi: %.4f <==|'
% (metrics.acc(self.y, y_pred), metrics.nmi(self.y, y_pred)))
def main():
x_train, y_train, x_test, y_test = get_data()
for n in [2, 3, 5, 10, 15]:
pca = decomposition.PCA(n_components=n)
pca.fit(x_train)
pca_x_train = pca.transform(x_train)
knn = KNeighborsClassifier(n_neighbors=3)
knn.fit(pca_x_train, y_train)
print('\nPCA: ', n)
y_train_pred = knn.predict(pca_x_train)
print('Training accuracy on selected features: %.3f' %
acc(y_train, y_train_pred))
pca_x_test = pca.transform(x_test)
y_test_pred = knn.predict(pca_x_test)
print('Testing accuracy on selected features: %.3f' %
acc(y_test, y_test_pred))
if n == 2:
fig, axs = plt.subplots(2)
fig.suptitle("PCA Scatter Plot", fontsize='small')
axs[0].scatter(pca_x_train[:, 0],
pca_x_train[:, 1],
marker='o',
c=y_train,
s=25,
edgecolor='k')
axs[1].scatter(pca_x_test[:, 0],
pca_x_test[:, 1],
marker='o',
c=y_test,
s=25,
edgecolor='k')
plt.show()
def cal_cost_tree(x, trn, trg):
x = [int(a) for a in np.round(x)]
if sum(x) == 0 : return np.inf
x_index = [i for i in range(len(x)) if x[i]==1]
trn = trn.reshape(trn.shape[1], -1)
trn = trn[x_index, :]
trn = np.transpose(trn)
clf = tree.DecisionTreeClassifier()
clf.fit(trn, trg)
pre = clf.predict(trn)
score = acc(pre, trg)
error = 1 - score
return (1-alpha)*error + alpha * (sum(x)*1.0/len(x)), error, sum(x)*1.0/len(x)
def updatePointsAtk(xc, clf, Xtr, ytr=None, Xtt=None, ytt=None, err_type='loss'):
d = Xtr.shape[1]
m = xc.size/d
xc = xc.reshape(m,d)
# if xcid.size != m:
# print 'Attack points indices do not match xc, exit!'
# return None
X0 = np.concatenate([Xtr, xc], axis=0)
# TODO: update SVC instead of retrain with xc
clf.fit(X0) # <---- this is just a lazy update
# maximize the obj value (generalization error)
if err_type is 'fval':
# objective values on untainted dataset Dtr
return 1*(clf.fval - clf.C*(clf.kdist2(xc) - clf.r).sum())
elif err_type is 'r':
# the squared radius
return 1*clf.r
elif err_type is 'xi':
return 1*(clf.fval - clf.C*(clf.kdist2(xc) - clf.r).sum() - clf.r)
elif err_type is 'f1':
if ytr is not None:
return f1_score(ytr, clf.y[:ytr.size])
else:
print 'You need give the true labels!'
return None
elif err_type is 'acc':
if Xtt is not None and ytt is not None:
y_clf = clf.predict_y(Xtt)
return acc(ytt, y_clf)
else:
print 'You need give the test dataset!'
return None
elif err_type is 'loss':
# min(\sum_xi - R^2) means let less samples lie out meanwhile maximize the ball
# note: xc are excluded
sum_xi_c = (clf.kdist2(xc) - clf.r).sum()
return (clf.fval - clf.r)/clf.C - sum_xi_c - clf.r
elif err_type is 'fn':
if ytr is not None:
pid = np.where(ytr==1)[0]
return 0.5*((ytr[pid] - clf.y[pid]).sum())/pid.size
else:
print 'You need give the true labels!'
return None
elif err_type is 'fp':
if ytr is not None:
nid = np.where(ytr==-1)[0]
return 0.5*((clf.y[nid] - ytr[nid]).sum())/nid.size
else:
print 'You need give the true labels!'
return None
def cal_cost_knn(x, trn, trg):
x = list(map(int, np.round(x)))
if sum(x) == 0 :
return np.inf, np.inf, 1
x_index = [i for i in range(len(x)) if x[i]==1]
trn = trn.reshape(trn.shape[1], -1)
trn = trn[x_index, :]
trn = np.transpose(trn)
clf = knn(n_neighbors=nn)
clf.fit(trn, trg)
pre = clf.predict(trn)
score = acc(pre, trg)
error = 1 - score
return (1-alpha)*error + alpha * (sum(x)*1.0/len(x)), error, sum(x)*1.0/len(x)
def cal_cost_svm(x, trn, trg):
x = [int(a) for a in np.round(x)]
if sum(x) == 0 :
return np.inf, np.inf, 1
x_index = [i for i in range(len(x)) if x[i]==1]
trn = trn.reshape(trn.shape[1], -1)
trn = trn[x_index, :]
trn = np.transpose(trn)
clf = SVC(gamma="auto", kernel=svm_kernel)
clf.fit(trn, trg)
pre = clf.predict(trn)
score = acc(pre, trg)
error = 1 - score
return (1-alpha)*error + alpha * (sum(x)*1.0/len(x)), error, sum(x)*1.0/len(x)
def cluster_acc(Y,clusterLabels):
import numpy as np
from collections import Counter
from sklearn.metrics import accuracy_score as acc
assert (Y.shape == clusterLabels.shape)
pred = np.empty_like(Y)
for label in set(clusterLabels):
mask = clusterLabels == label
sub = Y[mask]
target = Counter(sub).most_common(1)[0][0]
pred[mask] = target
# assert max(pred) == max(Y)
# assert min(pred) == min(Y)
return acc(Y,pred)
def dae_svm():
dae_train = np.load('data/train_dae.npy')[:10000]
dae_test = np.load('data/test_dae.npy')[:5000]
svm_dae = svm.SVC(C=2.0, gamma=0.05, cache_size=2000)
#svm_dae = GridSearchCV(svr, parameters)
svm_dae.fit(dae_train, label_train)
predicted_dae = svm_dae.predict(dae_test)
daeacc = acc(label_test, predicted_dae)
#model_params = str(svm_dae.best_estimator_)
print 'DAE accuracy - ' + str(daeacc)
with open(PATHS + 'dae_svm', 'wb') as f:
pickle.dump(daeacc, f)
def reportStats(weight, current_iteration, X_train, y_train, X_test, y_test):
y_train[y_train < 0] = 0
y_test[y_test < 0] = 0
ypred_is = predict_all(X_train, weight)
ypred_oos = predict_all(X_test, weight)
np_err_handling = np.seterr(invalid='ignore')
is_acc = acc(y_train, ypred_is)
is_mcc = mcc(y_train, ypred_is)
is_f1 = f1(y_train, ypred_is)
is_mse = mse(y_train, ypred_is)
oos_acc = acc(y_test, ypred_oos)
oos_mcc = mcc(y_test, ypred_oos)
oos_f1 = f1(y_test, ypred_oos)
oos_mse = mse(y_test, ypred_oos)
is_tn, is_fp, is_fn, is_tp = confusion_matrix(y_train, ypred_is).ravel()
oos_tn, oos_fp, oos_fn, oos_tp = confusion_matrix(y_test,
ypred_oos).ravel()
is_auprc = auprc(y_train, ypred_is)
oos_auprc = auprc(y_test, ypred_oos)
np.seterr(**np_err_handling)
print(
f"Consensus {current_iteration}: IS acc {is_acc:0.5f}. IS MCC {is_mcc:0.5f}. IS F1 {is_f1:0.5f}. IS MSE {is_mse:0.5f}. OOS acc {oos_acc:0.5f}. OOS MCC {oos_mcc:0.5f}. OOS F1 {oos_f1:0.5f}. OOS MSE {oos_mse:0.5f}."
)
print(
f"Confusion {current_iteration}: IS TP: {is_tp}, IS FP: {is_fp}, IS TN: {is_tn}, IS FN: {is_fn}, IS AUPRC: {is_auprc:0.5f}. OOS TP: {oos_tp}, OOS FP: {oos_fp}, OOS TN: {oos_tn}, OOS FN: {oos_fn}, OOS AUPRC: {oos_auprc:0.5f}."
)
return is_acc, is_mcc, is_f1, is_mse, is_auprc, oos_acc, oos_mcc, oos_f1, oos_mse, oos_auprc
def test_acc_knn(x, tst, tst_trg, trn, trn_trg):
x = [int(a) for a in np.round(x)]
if sum(x) == 0:
return 0
x = [i for i in range(len(x)) if x[i]==1]
tst = tst.reshape(tst.shape[1], -1)
tst = tst[x, :]
tst = np.transpose(tst)
trn = trn.reshape(trn.shape[1], -1)
trn = trn[x, :]
trn = np.transpose(trn)
clf = knn(n_neighbors=nn)
clf.fit(trn, trn_trg)
tst_pred = clf.predict(tst)
return acc(tst_trg, tst_pred)
def test_acc_svm(x, tst, tst_trg, trn, trn_trg):
x = [int(a) for a in np.round(x)]
if sum(x) == 0:
return 0
x = [i for i in range(len(x)) if x[i]==1]
tst = tst.reshape(tst.shape[1], -1)
tst = tst[x, :]
tst = np.transpose(tst)
trn = trn.reshape(trn.shape[1], -1)
trn = trn[x, :]
trn = np.transpose(trn)
clf = SVC(gamma="auto", kernel=svm_kernel)
clf.fit(trn, trn_trg)
tst_pred = clf.predict(tst)
return acc(tst_trg, tst_pred)
def test_acc_tree(x, tst, tst_trg, trn, trn_trg):
x = [int(a) for a in np.round(x)]
if sum(x) == 0:
return
x = [i for i in range(len(x)) if x[i]==1]
tst = tst.reshape(tst.shape[1], -1)
tst = tst[x, :]
tst = np.transpose(tst)
trn = trn.reshape(trn.shape[1], -1)
trn = trn[x, :]
trn = np.transpose(trn)
clf = tree.DecisionTreeClassifier()
clf.fit(trn, trn_trg)
tst_pred = clf.predict(tst)
return acc(tst_trg, tst_pred)
def cluster_acc(y, cluster_labels):
assert (y.shape == cluster_labels.shape)
pred = np.empty_like(
y
) # what the clustering algoritm predicts as the y-label (if it predicted the majority y-label for each cluster)
for label in set(cluster_labels):
mask = cluster_labels == label # indices where the training data match a specific cluster
sub = y[mask] # get only the training labels that are in this cluster
target = Counter(sub).most_common(1)[0][
0] # get the majority y-label from the instances assigned to this cluster (e.g 0)
pred[
mask] = target # assign the training data in this cluster to have the majority y-label
# assert max(pred) == max(Y)
# assert min(pred) == min(Y)
return acc(y, pred)
def get_accuracy(predictions, y, std_price = 0, mean_price = 0):
unnorm_predictions = []
for pred in predictions:
if math.isnan(unnormalize(pred, std_price,mean_price)):
print("NAN FOUND")
exit()
unnorm_predictions.append(unnormalize(pred, std_price,
mean_price))
unnorm_y = []
for y_pt in y:
if math.isnan(unnormalize(y_pt, std_price,mean_price)):
print("NAN FOUND")
exit()
unnorm_y.append(unnormalize(y_pt, std_price,
mean_price))
# Create lists to measure if opening price increased or decreased
direction_pred = []
for pred in unnorm_predictions:
if pred >= 0:
direction_pred.append(1)
else:
direction_pred.append(0)
direction_test = []
for value in unnorm_y:
if value >= 0:
direction_test.append(1)
else:
direction_test.append(0)
from sklearn.metrics import confusion_matrix
# Calculate if the predicted direction matched the actual direction
direction = acc(direction_test, direction_pred)
direction = round(direction,4)*100
_mae = mae(unnorm_y, unnorm_predictions) #median absolute error
_rmse = np.sqrt(mse(y, predictions)) # root mean squared error
_r2 = r2(unnorm_y, unnorm_predictions) #R squared error
print("CONFUSION MATRIX")
print(confusion_matrix(direction_test,direction_pred).ravel())
return (direction,_mae,_rmse,_r2)
def test(self, x_test, y_test, params, n_centers, width):
y_true = []
y_pred = []
(p, _) = x_test.shape
for i in range(p):
d = y_test[i]
y = self.predict(x_test[i], params, n_centers, width)
# Confusion Matrix
y_true.append(list(d))
y_pred.append(list(y))
a = util.inverse_transform(y_true, self.n_classes)
b = util.inverse_transform(y_pred, self.n_classes)
return acc(a, b), tpr(a, b,
average='macro'), 0, ppv(a,
b,
average='weighted')
#########################################################
### your code goes here ###
from sklearn.svm import SVC
from sklearn.metrics import accuracy_score as acc
for i in range(1,20):
C = pow(10,i)
clf = SVC(kernel="rbf",C=C)
clf.fit(features_train,labels_train)
pred = clf.predict(features_test)
print "C:",C,"Accuracy:",acc(pred,labels_test)
C = 10000
clf = SVC(kernel="rbf",C=C)
clf.fit(features_train,labels_train)
pred = clf.predict(features_test)
print "C:",C,"Accuracy:",acc(pred,labels_test)
from sklearn.tree import DecisionTreeClassifier as DTC
from sklearn.metrics import accuracy_score as accu
clf = DTC(min_samples_split=2)
clf.fit(features_train,labels_train)
pred = clf.predict(features_test)
features_train, features_test, labels_train, labels_test = cross_validation.train_test_split(word_data, authors, test_size=0.1, random_state=42)
for outcast in ["sshacklensf", "cgermannsf"]:
features_train = [x.replace(outcast,"") for x in features_train]
from sklearn.feature_extraction.text import TfidfVectorizer
vectorizer = TfidfVectorizer(sublinear_tf=True, max_df=0.5,
stop_words='english')
features_train = vectorizer.fit_transform(features_train)
features_test = vectorizer.transform(features_test).toarray()
### a classic way to overfit is to use a small number
### of data points and a large number of features;
### train on only 150 events to put ourselves in this regime
features_train = features_train[:150].toarray()
labels_train = labels_train[:150]
### your code goes here
from sklearn.tree import DecisionTreeClassifier
from sklearn.metrics import accuracy_score as acc
clf = DecisionTreeClassifier()
clf.fit(features_train,labels_train)
pred = clf.predict(features_test)
print acc(pred,labels_test)
print [(i,x) for i,x in enumerate(clf.feature_importances_) if x > 0.2]
print vectorizer.get_feature_names()[21323]
#### initial visualization
plt.xlim(0.0, 1.0)
plt.ylim(0.0, 1.0)
plt.scatter(bumpy_fast, grade_fast, color="b", label="fast")
plt.scatter(grade_slow, bumpy_slow, color="r", label="slow")
plt.legend()
plt.xlabel("bumpiness")
plt.ylabel("grade")
plt.show()
################################################################################
### your code here! name your classifier object clf if you want the
### visualization code (prettyPicture) to show you the decision boundary
from sklearn.ensemble import RandomForestClassifier
from sklearn.metrics import accuracy_score as acc
clf = RandomForestClassifier(n_jobs=-1, criterion="gini", n_estimators=100, min_samples_leaf=5, max_features=1)
clf.fit(features_train, labels_train)
pred = clf.predict(features_test)
print "Accuracy", acc(pred, labels_test)
try:
prettyPicture(clf, features_train, labels_train)
except NameError:
pass
