Линейная и Логистическая Регрессия

import numpy as np
import pandas as pd
from sklearn.datasets import load_breast_cancer 
from sklearn.model_selection import train_test_split as tts
from sklearn.preprocessing import normalize
 
#Загрузка данных
cancer = load_breast_cancer()
cancerdata = pd.DataFrame(np.c_[cancer['data'], cancer['target']], 
                          columns= np.append(cancer['feature_names'], ['target']))
cancerdata.head()
m, n = cancerdata.shape
X = cancerdata.iloc[:, :-1].values
Y = cancerdata.iloc[:, n-1].values
Y = Y.reshape(-1, 1)
train_X, test_X, train_Y, test_Y = tts(X, Y, test_size = 50/569, shuffle = False)
m1, n = train_X.shape
m2, n = test_X.shape
train_X = normalize(train_X)
test_X = normalize(test_X)
XT = train_X.T
 
#Основные функции
def sigmoid(z):
    return 1/(1+np.exp(-z))
 
def cross_entropy_loss(y_pred,target):
    return -np.mean((target*np.log(y_pred+0.000001)+(1-target)*np.log(1-y_pred+0.000001)))
 
#Treshold
def predict(X_test):
    preds = []
    for i in sigmoid(np.dot(X_test, W) + b):
        if i>0.5:
            preds.append(1)
        else:
            preds.append(0)
    return preds
 
np.random.seed(0)
W = np.random.uniform(0,1,size=(n,1))
b=0.5
 
 
for i in range(100000000):
    Z = np.dot(train_X, W) + b
    Y_output = sigmoid(Z)
    E = cross_entropy_loss(Y_output, train_Y)
    print("Cross entropy loss: ",E)
    #գրադիենտ
    grad = Y_output - train_Y
    grad_weight = np.dot(XT,grad)/m1
    grad_bias = np.average(grad)
    #alpha=0.01
    W = W-.01*grad_weight
    b = b-.01*grad_bias
 
pred_Y = np.array(predict(test_X))
 
pred_Y1 = []
for i in range(50):
    tmp = [pred_Y[i]]
    pred_Y1.append(tmp)
pred_Y = np.array(pred_Y1)
 
check = pred_Y - test_Y
unique, counts = np.unique(check, return_counts=True)

import numpy as np
import pandas as pd
from sklearn.datasets import load_boston
from sklearn.model_selection import train_test_split as tts
from sklearn.preprocessing import normalize
 
def init(n):
    return np.random.random((n,1))
 
boston = load_boston()
bostondata = pd.DataFrame(np.c_[boston['data'], boston['target']], 
                          columns= np.append(boston['feature_names'], ['target']))
X = bostondata.iloc[:, :-1].values
m, n = bostondata.shape
X = np.append(np.ones((m,1)), X, axis = 1)
Y = bostondata.iloc[:, n-1].values
Y = Y.reshape(-1, 1)
train_X, test_X, train_Y, test_Y = tts(X, Y, test_size = 106/506, shuffle = False)
m1, n = train_X.shape
m2, n = test_X.shape
train_X = normalize(train_X)
test_X = normalize(test_X)
 
teta = init(n)
XT = train_X.transpose()
dJ = np.dot(XT, (np.dot(train_X, teta)-train_Y))/m1
iteration = 0
while(np.max(np.abs(dJ))>0.00001):
    tetanew = teta-0.001*dJ
    teta = tetanew.copy()
    print(np.max(np.abs(dJ)))
    dJ = np.dot(XT, (np.dot(train_X, teta) - train_Y))/m1
    iteration += 1
 
pred_Y = np.dot(test_X, teta)
check = test_Y - pred_Y