def preprocess(self, data, size):
data = data.drop([3], axis=1)
rows, columns = data.shape
data[0] = np.ones(rows)
power = [i + 1 for i in range(columns - 1)] #1 to number of columns
val = 0
for i in range(2, size + 1): # 2 to size -- i
new_col = list(combinations_with_replacement(power, i))
for j in range(len(new_col)):
data[columns + val] = 1
for k in new_col[j]:
data[columns + val] = data[columns + val] * data[k]
val += 1
print("Data is of form in Data frame without Normalization: ")
print(data.head(5))
# n1 = MinMaxScaler() # normalizing data here
n1 = NormalScaler()
# n1.fit(data)
# data = n1.transform(data)
for i in data.columns:
n1.fit(data[i])
data[i] = n1.transform(data[i])
data[0] = 1
print("Data is of form in Data frame after Normalization: ")
print(data.head(5))
data = data.to_numpy()
print(data.shape)
return data
def preprocess(self, data, degree):
data = data.drop([3], axis=1)
rows, columns = data.shape
data[0] = np.ones(rows)
power = [i + 1 for i in range(columns - 1)]
val = 0
for i in range(2, degree + 1): # i range from 2 to degree
new_col = list(combinations_with_replacement(power, i))
for j in range(len(new_col)):
data[columns + val] = 1
for k in new_col[j]:
data[columns + val] = data[columns + val] * data[k]
val += 1
# print("Data is of form in Data frame without Normalization: ")
# print(data.head(2))
# n1 = MinMaxScaler()
n1 = NormalScaler()
n1.fit(data)
data = n1.transform(data)
data[0] = 1
# print("Data is of form in Data frame after Normalization: ")
# print(data.head(2))
data = data.to_numpy()
print("Shape of Data is :")
print(data.shape)
return data
def preprocess(self, data, size):
data = data.drop([3], axis=1)
rows, columns = data.shape
data[0] = np.ones(rows)
power = [i + 1 for i in range(columns - 1)] #1 to number of columns
val = 0
for i in range(2, size + 1): # 2 to size -- i
new_col = list(combinations_with_replacement(power, i))
for j in range(len(new_col)):
data[columns + val] = 1
for k in new_col[j]:
data[columns + val] = data[columns + val] * data[k]
val += 1
n1 = NormalScaler()
for i in data.columns:
n1.fit(data[i])
data[i] = n1.transform(data[i])
data[0] = 1
data = data.to_numpy()
return data
"""
X = self.add_bias(X)
return np.matmul(X, self.W)
if __name__ == "__main__":
model = LinearRegression_Vectorized()
# data input
data = pd.read_csv("./3D_spatial_network.csv", header=None)[::]
X = data.loc[:, 0:1].values
y = data.loc[:, 2].values
# data preprocessing (Normal Scaling)
mscaler = NormalScaler()
mscaler.fit(X[:, 0])
X[:, 0] = mscaler.transform(X[:, 0])
mscaler.fit(X[:, 1])
X[:, 1] = mscaler.transform(X[:, 1])
# training the model
arr = model.train(X, y)
print("weights: ", model.W)
print("Total Cost: ", model.cost)
# visualization of cost function.
res = 100
bounds = 10
xx = np.linspace(model.W[1] - bounds, model.W[1] + bounds, res)
yy = np.linspace(model.W[2] - bounds, model.W[2] + bounds, res)
minw0 = model.W[0]
from elm import ELM
from MLP_auto import MLP
if __name__ == '__main__':
data = pd.DataFrame(loadmat('./data5.mat')['x'])
data = data.sample(frac=1).reset_index(drop=True)
X = data.loc[:, :71].values
y = data.loc[:, 72:73].values
y_cat = np.zeros((y.shape[0], 2))
for i in range(y.shape[0]):
y_cat[i][int(y[i])] = 1
# data preprocessing
scaler = NormalScaler()
for j in range(X.shape[1]):
scaler.fit(X[:, j])
X[:, j] = scaler.transform(X[:, j])
# m = number of feature vectors
m = X.shape[0]
# n = number of features
n = X.shape[1]
train_percent = 0.6
X_train = X[:int(train_percent * X.shape[0]), :]
y_train = y_cat[:int(train_percent * X.shape[0]), :]
X_test = X[int(train_percent * X.shape[0]):, :]
y_test = y_cat[int(train_percent * X.shape[0]):, :]
alpha = 0.6
max_iter = 25
from preprocessing import NormalScaler
import matplotlib.pyplot as plt
import keras
# loading data
data = loadmat('./data_for_cnn.mat')['ecg_in_window'].astype(np.float64)
data_labels = loadmat('./class_label.mat')['label'].astype(np.int)
data = np.concatenate((data, data_labels), axis=1)
np.random.shuffle(data)
# data preprocessing
scaler = NormalScaler()
for j in range(data.shape[1] - 1):
scaler.fit(data[:, j])
data[:, j] = scaler.transform(data[:, j])
# splitting data into train and test sets
split_percent = 0.8
X_train = data[:int(data.shape[0] * split_percent), :1000].astype(np.float)
y_train = data[:int(data.shape[0] * split_percent), 1000:1001]
X_test = data[int(data.shape[0] * split_percent):, :1000].astype(np.float)
y_test = data[int(data.shape[0] * split_percent):, 1000:1001]
X_train = X_train.reshape(X_train.shape[0], 1000, 1)
X_test = X_test.reshape(X_test.shape[0], 1000, 1)
# Convolutinal Neural Network model
model = Sequential()
if __name__ == "__main__":
# data input
data = pd.read_excel("./data4.xlsx", header=None)
data = data.sample(frac=1).reset_index(drop=True)
data = data.values
X = data[:, :7]
y = data[:, 7]
unique_classes = np.unique(y)
num_classes = len(unique_classes)
# data preprocessing
mscaler = NormalScaler()
for j in range(X.shape[1]):
mscaler.fit(X[j])
X[j] = mscaler.transform(X[j])
y_cat = (y == unique_classes[0]).astype('int').reshape(-1, 1)
for i in unique_classes[1:]:
y_cat = np.concatenate((y_cat, (y == i).astype('int').reshape(-1, 1)),
axis=1)
# splitting data using holdout cross validation
train_percent = 0.7
X_train = X[:int(train_percent * X.shape[0])]
y_train = y[:int(train_percent * X.shape[0])]
y_cat_train = y_cat[:int(train_percent * X.shape[0])]
X_test = X[int(train_percent * X.shape[0]):]
y_test = y[int(train_percent * X.shape[0]):]
y_cat_test = y_cat[int(train_percent * X.shape[0]):]
if __name__ == "__main__":
model = LinearRegression()
# data input
data = pd.read_csv("./3D_spatial_network.csv", header=None)[::]
data = data.sample(frac=1)
X = data.loc[:, 0:1].values
y = data.loc[:, 2].values
X = get_polyfeatures(X, 6)
# data preprocessing (Normal scaling)
mscaler = NormalScaler()
for j in range(X.shape[1]):
mscaler.fit(X[:, j])
X[:, j] = mscaler.transform(X[:, j])
# X = get_polyfeatures(X, 6)
train_percent = 0.6
cv_percent = 0.2
test_percent = 1 - train_percent - cv_percent
X_train = X[:int(train_percent * X.shape[0])]
y_train = y[:int(train_percent * X.shape[0])]
X_cv = X[int(train_percent * X.shape[0]):int((train_percent + cv_percent) *
X.shape[0])]
y_cv = y[int(train_percent * X.shape[0]):int((train_percent + cv_percent) *
X.shape[0])]
X_test = X[int((train_percent + cv_percent) * X.shape[0]):]
return y_pred
if __name__ == "__main__":
model = LogisticRegression()
# data input
data = pd.read_excel("./data3.xlsx", header=None)
data = data.sample(frac=1).reset_index(drop=True)
X = data[[0, 1, 2, 3]]
y = data[4] - 1
# data preprocessing (Normal scaling)
mscaler = NormalScaler()
for j in range(X.shape[1]):
mscaler.fit(X.loc[:, j])
X.loc[:, j] = mscaler.transform(X.loc[:, j])
# holdout cross validation split
train_percent = 0.6
X_train = X[:int(train_percent * X.shape[0])]
y_train = y[:int(train_percent * X.shape[0])]
X_test = X[int(train_percent * X.shape[0]):]
y_test = y[int(train_percent * X.shape[0]):]
# Training the model by choosing alpha and max_iter values.
# gradient descent algorithm can be set as either ‘batch’ or ‘stochastic’
# in this function call.
alpha = 0.26
max_iter = 100
algo = 'batch'