def train(self, train_data, train_label):
if self.norm_type == "Standardization":
train_data = preProcess.Standardization(train_data)
else:
train_data = preProcess.Normalization(train_data)
skf = StratifiedKFold(self.n_folds)
prediction_feature = np.zeros(
(train_data.shape[0], len(self.classifier_set)))
trained_model = []
# the first layer in Stacking
for j, clf in enumerate(self.classifier_set):
# train each submodel
subtrained_model = []
# cross validation
for (train_index, test_index) in skf.split(train_data,
train_label):
X_train, X_test = train_data[train_index], train_data[
test_index]
y_train, y_test = train_label[train_index], train_label[
test_index]
# train and save the model trained with S-si
clf.train(X_train, y_train)
subtrained_model.append(clf)
# get the prediction feature for each sub model
prediction_feature[test_index, j] = clf.predict(X_test)[:, 0]
# save the models
trained_model.append(subtrained_model)
self.trained_classifier_set = trained_model
return self
def train(self, train_data, train_label, method="GA", alpha=0.1, iterations=100):
if self.norm_type == "Standardization":
train_data = preProcess.Standardization(train_data)
else:
train_data = preProcess.Normalization(train_data)
train_label = np.expand_dims(train_label, axis=1)
feature_dim = len(train_data[1])
if method == "GA":
weights = np.random.normal(0, 1, [feature_dim, 1])
for i in range(iterations):
pred = self.sigmoid(np.dot(train_data, weights))
errors = train_label - pred
# update the weights
weights = weights + alpha * np.dot(train_data.T, errors)
self.weights = weights
return self
if method == "SGA":
weights = np.random.normal(0, 1, feature_dim)
sample_num = len(train_data)
random_index = np.random.randint(sample_num, size=sample_num)
for i in range(iterations):
for j in range(sample_num):
alpha = self.updataAlpha(alpha, i, 1)
pred = self.sigmoid(np.dot(train_data[random_index[j], :], weights))
sample_error = train_label[random_index[j]] - pred
weights = weights + alpha * sample_error * train_data[random_index[j], :]
self.weights = weights
return self
def predict(self, test_data, prob="False"):
# Normalization
if self.norm_type == "Standardization":
test_data = preProcess.Standardization(test_data)
else:
test_data = preProcess.Normalization(test_data)
test_num = test_data.shape[0]
prediction = np.zeros([test_num, 1])
probability = np.zeros([test_num, 1])
for i in range(test_num):
inter_1 = test_data[i] * self.V
inter_2 = np.multiply(test_data[i], test_data[i]) * np.multiply(self.V, self.V)
interaction = sum(np.multiply(inter_1, inter_1) - inter_2) / 2.
pre = self.w_0 + test_data[i] * self.W + interaction
probability = self.sigmoid(float(pre))
if probability[i] > 0.5:
prediction[i] = 1
else:
prediction[i] = 0.5
self.prediction = prediction
self.probability = probability
if prob:
return probability
else:
return prediction
def predict(self, test_data, prob="False"):
# Normalization
if self.norm_type == "Standardization":
test_data = preProcess.Standardization(test_data)
else:
test_data = preProcess.Normalization(test_data)
test_num = test_data.shape[0]
prediction = np.zeros([test_num, 1])
probability = np.zeros([test_num, 1])
# find the support vectors and its corresponding label
support_vectors_index = np.nonzero(self.alphas > 0)[0]
support_vectors = self.train_data[support_vectors_index]
support_vectors_label = self.train_label[support_vectors_index]
support_vectors_alphas = self.alphas[support_vectors_index]
# predict the test sample in page of 122 Eq.(7.89)
for i in range(test_num):
kernel_data = self.kernelTransformation(support_vectors, test_data[i, :], self.kernel)
probability[i] = np.dot(kernel_data.T, np.multiply(support_vectors_label, support_vectors_alphas)) + self.b
if probability[i] > 0:
prediction[i] = 1
else:
prediction[i] = -1
self.prediction = prediction
self.probability = probability
if prob:
return probability
else:
return prediction
def train(self, train_data, train_label):
if self.norm_type == "Standardization":
train_data = preProcess.Standardization(train_data)
else:
train_data = preProcess.Normalization(train_data)
# initiation
sample_num, feature_dim = np.shape(train_data)
self.train_data = train_data
self.train_label = self.labelTransformation(train_label)
self.sample_num = sample_num
self.K = np.zeros([self.sample_num, self.sample_num])
self.alphas = np.zeros([self.sample_num, 1])
self.errors = np.zeros([self.sample_num, 2])
self.b = 0
# kernel trick
for i in range(self.sample_num):
self.K[:, i] = self.kernelTransformation(self.train_data,
self.train_data[i, :],
self.kernel)
# train model
self.SMO()
return self
def train(self, train_data, train_label, alpha=0.01, iterations=100):
if self.norm_type == "Standardization":
train_data = preProcess.Standardization(train_data)
else:
train_data = preProcess.Normalization(train_data)
for epoch in range(iterations):
for id in range(self.sample_num):
# second order computation
inter_1 = train_data[id] * self.V
inter_2 = np.multiply(train_data[id], train_data[id]) * np.multiply(self.V, self.V)
interaction = np.sum(np.multiply(inter_1, inter_1) - inter_2) / 2.
# prediction result
pred = self.w_0 + train_data[id] * self.W + interaction
# calculate loss, cross entropy
base = [np.log(self.sigmoid(train_label[id] * float(pred))) - 1] * train_label
# update numerical parameters
self.w_0 -= alpha * base
x = train_data[id]
for i in range(self.n):
# update first-order parameter
if train_data[id, i] != 0:
self.W[id, i] -= alpha * base * train_data[id, i]
for j in range(self.n):
# update second-order parameter
self.V[i, j] -= alpha * base * (
train_data[id, i] * self.V[j, i] * train_data[id, j] - self.V[i, j] * train_data[id, i] * train_data[id, i])
return self
def train(self, train_data, train_label):
if self.norm_type == "Standardization":
train_data = preProcess.Standardization(train_data)
else:
train_data = preProcess.Normalization(train_data)
train_data1, train_data2, train_label1, train_label2 = train_test_split(
train_data, train_label, test_size=0.5, random_state=2019)
# train set in the second layer
train_predict_feature = np.zeros((train_data2.shape[0], self.k))
trained_model = []
# the first layer in Blending
for j, clf in enumerate(self.classifier_set):
# train each submodel
print(j, clf)
clf.train(train_data1, train_label1)
train_predict_feature[:, j] = clf.predict(train_data2)[:, 0]
# save the trained model in the first layer
trained_model.append(clf)
# the second layer in Blending
layer2_clf = PerceptronClassifier()
layer2_clf.train(train_predict_feature, train_label2)
self.layer1_classifier_set = trained_model
self.layer2_classifier = layer2_clf
return self
def predict(self, test_data, prob="False"):
# Normalization
if self.norm_type == "Standardization":
test_data = preProcess.Standardization(test_data)
else:
test_data = preProcess.Normalization(test_data)
test_num = test_data.shape[0]
prediction = np.zeros([test_num, 1])
probability = np.zeros([test_num, 1])
for i in range(test_num):
probability[i] = self.sigmoid(
np.dot(self.w.T, test_data[i, :]) + self.b
) # prediction = self.sigmoid(np.dot(self.w.T, test_data) + self.b) can speed up
if probability[i] > 0:
prediction[i] = 1
else:
prediction[i] = -1
self.prediction = prediction
self.probability = probability
if prob:
return probability
else:
return prediction
def train(self, train_data, train_label):
if self.norm_type == "Standardization":
train_data = preProcess.Standardization(train_data)
else:
train_data = preProcess.Normalization(train_data)
self.x_train = train_data
self.y_train = train_label
return self
def cluster(self, train_data, display="True"):
if self.norm_type == "Standardization":
train_data = preProcess.Standardization(train_data)
else:
train_data = preProcess.Normalization(train_data)
sample_num = len(train_data)
label = -np.ones([sample_num])
center_flag = np.zeros([sample_num
]) # indicate samples which are cluster center
center_index = [] # cluster center index
C = 0
# start clustering
for i in range(sample_num):
temp_neighbor = []
if label[i] != -1: # if sample i has been clustered, stop this iteration
continue
index1, neighbor1 = self.neighborQuery(train_data,
train_data[i, :])
if len(neighbor1) < self.m:
label[i] = 0 # sample i is noise point
continue # stop this iteration
C = C + 1
label[i] = C
center_flag[i] = 1
temp_neighbor.append(neighbor1)
# all the samples in i-th neighbor belong to label C
center_index = index1
j = 0
while j < len(center_index):
loc = center_index[j]
#print(center_index[j])
if center_flag[
j] != 1: # samples in i-th sample's neighbor but not cluster center
if label[loc] == 0:
label[loc] = C
if label[loc] != -1:
j = j + 1
continue
label[loc] = C
index2, neighbor2 = self.neighborQuery(
train_data, train_data[loc, :])
if len(neighbor2) >= self.m:
center_index = list(
set(index1).union(set(index2))
) # merge the cluster centers which are directly reachable
center_flag[
loc] = 1 # if there are more than m samples in j-th neighbor, add j-th sample as a cluster center
j = j + 1
self.label = label
if display:
self.plotResult(train_data)
return label
def predict(self, x, prob="False"):
# Normalization
if self.norm_type == "Standardization":
x = preProcess.Standardization(x)
else:
x = preProcess.Normalization(x)
y = np.dot(x, self.w)
self.prediction = y
return y
def standardLinearRegression(self, x, y):
if self.norm_type == "Standardization":
x = preProcess.Standardization(x)
else:
x = preProcess.Normalization(x)
xTx = np.dot(x.T, x)
if np.linalg.det(xTx) == 0: # calculate the Determinant of xTx
print("Error: Singluar Matrix !")
return
w = np.dot(np.linalg.inv(xTx), np.dot(x.T, y))
return w
def lassoRegression(self, x, y):
if self.norm_type == "Standardization":
x = preProcess.Standardization(x)
else:
x = preProcess.Normalization(x)
sample_num, feataure_dim = np.shape(x)
w = np.zeros([feataure_dim, 1])
for i in range(self.iterations):
last_w = w
w[i] = np.dot(x[i, :], (y[i] - x[i, :] * last_w.T))/np.dot(x[i, :], x[i, :].T)
return w
def train(self, trainData, trainLabel):
if self.norm_type == "Standardization":
trainData = preProcess.Standardization(trainData)
else:
trainData = preProcess.Normalization(trainData)
trainLabel = np.expand_dims(trainLabel, axis=1)
data = np.hstack([trainData, trainLabel])
self.tree_node = self.createDecisionTree(data)
#self.printTree(self.tree_node)
return self
def train(self, data, label):
# Normalization
if self.norm_type == "Standardization":
data = preProcess.Standardization(data)
else:
data = preProcess.Normalization(data)
unique_label = np.unique(label)
mu = np.mean(data, axis=0)
# St = np.dot((data - mu).T, data - mu)
Sw = 0
Sb = 0
for c in unique_label:
index = np.where(label == c)
Ni = len(index)
xi = data[index]
mui = np.mean(xi, axis=0)
# calculate Sw
Si = np.dot((xi - mui).T, xi - mui)
Sw = Sw + Si
# calculate Sb
delta = np.expand_dims(mu - mui, axis=1)
Sb = Sb + Ni * np.dot(delta, delta.T)
# calculate the eigenvalue, eigenvector of Sw-1 * Sb
temp = np.dot(np.linalg.inv(Sw), Sb)
eigenvalue, eigenvector = np.linalg.eig(np.dot(np.linalg.inv(Sw), Sb))
index = np.argsort(-eigenvalue)
eigenvalue = eigenvalue[index]
eigenvector = eigenvector[:, index]
# calculate contribute rate
contribute_rate = np.zeros(len(index))
acc_contribute_rate = np.zeros(len(index))
value_sum = eigenvalue.sum()
sum = 0
k = 0
for i in range(len(eigenvalue)):
sum = sum + eigenvalue[i]
contribute_rate[i] = eigenvalue[i] / value_sum
acc_contribute_rate[i] = sum / value_sum
if (acc_contribute_rate[i - 1] <
self.rate) and (acc_contribute_rate[i] >= self.rate):
k = i
self.contribute_rate = contribute_rate
self.acc_contribute_rate = acc_contribute_rate
matrix = np.mat(eigenvector)[:, k]
self.matrix = matrix
return self
def train(self, train_data, train_label):
if self.norm_type == "Standardization":
train_data = preProcess.Standardization(train_data)
else:
train_data = preProcess.Normalization(train_data)
label_count = {}
feature_dim = len(train_data[1])
# get the number of each labels
for c in train_label:
label_count[c] = label_count.get(c, 0) + 1
label_value = sorted(label_count.items(), key=op.itemgetter(0), reverse=False)
self.label_value = label_value
K = len(label_value) # the number of unique labels
N = len(train_label) # the number of samples
# get the prior probability
prior_probability = {}
for key in range(len(label_value)):
prior_probability[label_value[key][0]] = (label_value[key][1] + self.laplace) / (N + K * self.laplace) # laplace smooth
self.prior_probability = prior_probability
# get the value set of each feature
feature_value = [] # feature with different value
S = [] # the number of unique values of each feature
for feat in range(feature_dim):
unique_feature = np.unique(train_data[:, feat])
S.append(len(unique_feature))
feature_value.append(unique_feature)
self.S = S
self.feature_value = feature_value
# calculate the conditional probability
prob = []
# calculate the count (x = a & y = c)
for j in range(feature_dim):
count = np.zeros([S[j], len(label_count)]) # the range of label start with 1
feature_temp = train_data[:, j]
feature_value_temp = feature_value[j]
for i in range(len(feature_temp)):
for k in range(len(feature_value_temp)):
for t in range(len(label_count)):
if feature_temp[i] == feature_value_temp[k] and train_label[i] == label_value[t][0]:
count[k][t] += 1 # x = value and y = label
# calculate the conditional probability
for m in range(len(label_value)):
count[:, m] = (count[:, m] + self.laplace) / (label_value[m][1] + self.laplace*S[j]) # laplace smoothing
# print(count)
prob.append(count)
self.conditional_probability = prob
return self
def ridgeRegression(self, x, y):
if self.norm_type == "Standardization":
x = preProcess.Standardization(x)
else:
x = preProcess.Normalization(x)
feature_dim = len(x[0])
xTx = np.dot(x.T, x)
matrix = xTx + np.exp(feature_dim)*self.lamda
if np.linalg.det(xTx) == 0:
print("Error: Singluar Matrix !")
return
w = np.dot(np.linalg.inv(matrix), np.dot(x.T, y))
return w
def LWLinearRegression(self, x, y, sample):
if self.norm_type == "Standardization":
x = preProcess.Standardization(x)
else:
x = preProcess.Normalization(x)
sample_num = len(x)
weights = np.eye(sample_num)
for i in range(sample_num):
diff = sample - x[i, :]
weights[i, i] = np.exp(np.dot(diff, diff.T)/(-2 * self.k ** 2))
xTx = np.dot(x.T, np.dot(weights, x))
if np.linalg.det(xTx) == 0:
print("Error: Singluar Matrix !")
return
result = np.dot(np.linalg.inv(xTx), np.dot(x.T, np.dot(weights, y)))
return result
def cluster(self, train_data, display="True"):
if self.norm_type == "Standardization":
train_data = preProcess.Standardization(train_data)
else:
train_data = preProcess.Normalization(train_data)
if self.cluster_type == "KMeans":
self.centers, self.distances = self.kmeans(train_data, self.k)
elif self.cluster_type == "biKMeans":
self.centers, self.distances = self.biKmeans(train_data)
elif self.cluster_type == "KMeans++":
self.centers, self.distances = self.kmeansplusplus(train_data)
else:
print("Wrong cluster type!")
sys.exit()
if display:
self.plotResult(train_data)
return self.distances[:, 0]
def predict(self, test_data):
# Normalization
if self.norm_type == "Standardization":
testData = preProcess.Standardization(test_data)
else:
testData = preProcess.Normalization(test_data)
test_num = testData.shape[0]
prediction = np.zeros([test_num, 1])
probability = np.zeros([test_num, 1])
# predict each samples in test data
for i in range(test_num):
prediction[i], probability[i] = self.calculateDistance(
testData[i], self.x_train, self.y_train, self.k)
self.prediction = prediction
self.probability = probability
return prediction
def train(self, train_data, train_label):
if self.norm_type == "Standardization":
train_data = preProcess.Standardization(train_data)
else:
train_data = preProcess.Normalization(train_data)
if self.regression_type == "Standard":
self.w = self.standardLinearRegression(train_data, train_label)
elif self.regression_type == "Localweight":
self.w = self.LWLinearRegression(train_data, train_label)
elif self.regression_type == "Ridge":
self.w = self.ridgeRegression(train_data, train_label)
elif self.regression_type == "Lasso":
self.w = self.lassoRegression(train_data, train_label)
elif self.regression_type == "Forwardstep":
self.w = self.forwardstepRegression(train_data, train_label)
else:
print("Error Regression Type!")
return self
def train(self, train_data, train_label):
if self.norm_type == "Standardization":
train_data = preProcess.Standardization(train_data)
else:
train_data = preProcess.Normalization(train_data)
train_label = np.expand_dims(train_label, axis=1)
sample_num = len(train_data)
weak_classifier = []
# initialize weights
w = np.ones([sample_num, 1])
w = w / sample_num
# predictions
agg_predicts = np.zeros([sample_num,
1]) # aggregate value of prediction
# start train
for i in range(self.iterations):
base_clf, error, base_prediction = self.baseClassifier(
train_data, train_label, w)
alpha = self.updateAlpha(error)
weak_classifier.append((alpha, base_clf))
# update parameters in page of 139 Eq.(8.4)
expon = np.multiply(-1 * alpha * train_label, base_prediction)
w = np.multiply(w, np.exp(expon))
w = w / w.sum()
# calculate the total error rate
agg_predicts += alpha * base_prediction
error_rate = np.multiply(
np.sign(agg_predicts) != train_label, np.ones([sample_num, 1]))
error_rate = error_rate.sum() / sample_num
if error_rate == 0:
break
self.classifier_set = weak_classifier
return weak_classifier
def predict(self, test_data, prob="False"):
# Normalization
if self.norm_type == "Standardization":
test_data = preProcess.Standardization(test_data)
else:
test_data = preProcess.Normalization(test_data)
test_num = test_data.shape[0]
prediction = np.zeros([test_num, 1])
probability = np.zeros([test_num, 1])
for i in range(test_num):
result = self.classify(test_data[i, :], self.tree_node)
result = sorted(result.items(), key=op.itemgetter(1), reverse=True)
prediction[i] = result[0][0]
#probability[i] = result[0][1]/(result[0][1] + result[1][1])
self.prediction = prediction
self.probability = probability
if prob:
return probability
else:
return prediction
def predict(self, test_data, prob="False"):
# Normalization
if self.norm_type == "Standardization":
test_data = preProcess.Standardization(test_data)
else:
test_data = preProcess.Normalization(test_data)
test_predict_feature = np.zeros((test_data.shape[0], self.k))
# the first layer in Blending
for j, clf in enumerate(self.layer1_classifier_set):
test_predict_feature[:, j] = clf.predict(test_data)[:, 0]
# the second layer in Blending
probability = self.layer2_classifier.predict(test_predict_feature)
prediction = (probability > 0.5) * 1
self.probability = probability
self.prediction = prediction
if prob:
return probability
else:
return prediction
def forwardstepRegression(self, x, y):
if self.norm_type == "Standardization":
x = preProcess.Standardization(x)
else:
x = preProcess.Normalization(x)
sample_num, feature_dim = np.shape(x)
w = np.zeros([self.iterations, feature_dim])
best_w = np.zeros([feature_dim, 1])
for i in range(self.iterations):
min_error = np.inf
for j in range(feature_dim):
for sign in [-1, 1]:
temp_w = best_w
temp_w[j] += sign * self.learning_rate
y_hat = np.dot(x, temp_w)
error = ((y - y_hat) ** 2).sum() # MSE
if error < min_error: # save the best parameters
min_error = error
best_w = temp_w
w[i, :] = best_w.T
return w
def predict(self, test_data, prob="False"):
# Normalization
if self.norm_type == "Standardization":
test_data = preProcess.Standardization(test_data)
else:
test_data = preProcess.Normalization(test_data)
test_num = test_data.shape[0]
prediction = np.zeros([test_num, 1])
probability = np.zeros([test_num, 1])
for classifier in self.classifier_set:
alpha = classifier[0]
clf = classifier[1]
base_prediction = alpha * clf.predict(test_data)
probability += base_prediction
self.prediction = np.sign(probability)
self.probability = probability
if prob:
return probability
else:
return prediction
def predict(self, test_data, prob="False"):
# Normalization
if self.norm_type == "Standardization":
test_data = preProcess.Standardization(test_data)
else:
test_data = preProcess.Normalization(test_data)
pre_prediction = np.zeros((test_data.shape[0], self.n_folds))
# the first layer in Stacking
for j, sub_model in enumerate(self.trained_classifier_set):
sub_prediction_feature = np.zeros(
(test_data.shape[0], self.n_folds))
i = 0
for clf in sub_model:
sub_prediction_feature[:, i] = clf.predict(test_data)[:, 0]
i = i + 1
pre_prediction[:, j] = sub_prediction_feature.mean(1)
test_num = test_data.shape[0]
prediction = np.zeros([test_num, 1])
probability = np.zeros([test_num, 1])
# the second layer in Stacking
if self.fusion_type == "Averaging":
probability = pre_prediction.mean(1)
elif self.fusion_type == "Voting":
probability = np.sum(pre_prediction, axis=1) / self.k
elif self.fusion_type == "Weighing":
w = [i / i.sum() for i in pre_prediction]
probability = np.sum(np.multiply(pre_prediction, w), axis=1)
prediction = (probability > 0.5) * 1
self.probability = probability
self.prediction = prediction
if prob:
return probability
else:
return prediction
def train(self, train_data, train_label):
if self.norm_type == "Standardization":
train_data = preProcess.Standardization(train_data)
else:
train_data = preProcess.Normalization(train_data)
feature_dim = len(train_data[1])
train_label = np.expand_dims(train_label, axis=1)
self.initializeParameter(feature_dim)
self.loss = []
# training process
for i in range(self.iterations):
gradients, cost = self.backPropagate(train_data, train_label)
# get the derivative
dw = gradients["dw"]
db = gradients["db"]
# update parameter
self.w = self.w - self.learning_rate * dw
self.b = self.b - self.learning_rate * db
self.loss.append(cost)
return self