3.

Logistic Regression

● Logistic regression is the most fundamental algorithm for handling binary

classification problems
● It's a commonly used algorithm for constructing a baseline model, and is generally a
good first model to build, because it's highly interpretable
● Logistic regression is similar to linear regression
○ We're still dealing with the line equation for making predictions, but
this time the result is passed through the sigmoid function to turn
ouptut into the probability
○ The probability tells you the chance of this instance belonging to
the positive classs
■ For example, if you have two classes (bad, good),
"good" is the positive class, and output of 0.8
means the model is 80% condifdent the instance
belongs to the positive class
● In a nutshell, logistic regression model outputs probabilities instead of raw classes
○ We then get the classes by applying a certain threshold
○ For example, if the probability is greater than 0.5, we assign the
positive class, and negtive class in the other case
○ This threshold can (and should) be altered depending on the
problem and the type of metric you're optimizing for
○ More on that later

● Logistic regression DOES NOT assume:

○ Linear relationship between the dependent and independent
variables
○ Normal distribution of the residuals
○ Homoscedasticity - constant variance in the residuals
● Logistic regression DOES assume:
○ Independent observations
○ Little or no multicollinearity among independent variables
○ Linearity of indenpendent variables and log odds
 ○ Large sample size - minimum of 10 cases with the least frequent
outcome for each independent variable in your model
■ E.g., if you have 5 independent variables and
expected probability of the least frequenct
outcome is 0.1, then you need a minimum sample
size of 500 (10 * 5 / 0.1)

Math behind

● In a way, we're still dealing with a line equation

● The formula of a line equation

● The Sigmoid/Logistic function is applied to the result

● The formula for this function is as follows:

● Let's visualize the function to see what happens behind the scenes
In [1]:
import numpy as np
import matplotlib.pyplot as plt
from matplotlib import rcParams
rcParams['figure.figsize'] = (14, 7)
rcParams['axes.spines.top'] = False
rcParams['axes.spines.right'] = False

def sigmoid(x):
return 1 / (1 + np.exp(-x))

xs = np.arange(-10, 10, 0.1)

ys = [sigmoid(x) for x in xs]

plt.plot(xs, ys, c='#087E8B', lw=3)

plt.title('Sigmoid function', size=20)
plt.xlabel('X', size=14)
plt.ylabel('Y (probability)', size=14)
plt.savefig('images/003_LogisticRegression_sigmoid_function.png', dpi=300, bbox_inches='tight')
plt.show()

● Essentially, sigmoid function returns a value between 0 and 1

● This value is interpreted as a probability of the positive class
● We can make classification from probabilities
○ If the probability is greater than some threshold (commonly 0.5),
assign the positive class
○ If the probability is less than some threshold (commonly 0.5),
assign the negative class
● As with linear regression, there are two parameters we need to optimize for -
weights and bias
● To perform the optimization, we'll need to declare Cost function
○ We can't use Mean Squared Error as with linear regression
○ We can, in theory, but it's not a good idea, because MSE was
designed to evaluate regression models
○ Instead, we can use the Cross Entropy function
● Formula of the Binary Cross Entropy function (2 classes):

● The in the formula can be re-written as a sigmoid function for a given input
● A brief introduction to BCE is made later
● Next you'll need to use this cost function in gradient descent to iteratively update
weights and bias
● To do so, you'll have to calculate partial derivatives of the cross entropy function
wrt. weights and bias parameters:

● This can be ommited, or you can leave it - it's not important

● Next, we're updating the existing weights and bias according to the following
formulas:

● Where is the learning rate

● This process is then repeated for a predefined number of iterations
 ● Before we see this in action, let's make a brief introduction to the Binary Cross
Entropy function

A Brief Introduction to Binary Cross Entropy

● Binary cross entropy function is a common cost function for evaluating binary
classification machine learning algorithms
● Commonly refered to as log loss
● It's value can be calculated from the previously mentioned formula
● This cost function "punishes" wrong predictions much more than it "rewards" good
predictions

Example 1: Calculating BCE for a correct prediction

● Let's say your model predicts the positive class with 90% probability
● This means the model is only 10% confident the negative class should be predicted
● What's the cross-entropy value?

● As you can see, the value of cross-entropy (or loss) is rather small

Example 2: Calculating BCE for an incorrect prediction

● Let's say your model predicts the positive class with 10% probability
● This means the model is 90% confident the negative class should be predicted
● What's the cross-entropy value?

● As you can see, the loss here is quite big

● We can calculate both with Python:
In [2]:
def binary_cross_entropy(y, y_hat):
def safe_log(x): return 0 if x == 0 else np.log(x)

total = 0
 for curr_y, curr_y_hat in zip(y, y_hat):
total += (curr_y * safe_log(curr_y_hat) + (1 - curr_y) * safe_log(1 - curr_y_hat))
return - total / len(y)
● We need the safe_log() function inside because log(0) returns infinity:
In [3]:
np.log(0)

● Let's evalutate for Example 1:

In [4]:
binary_cross_entropy(y=[1, 0], y_hat=[0.9, 0.1])

● And now for Example 2:

In [5]:
binary_cross_entropy(y=[1, 0], y_hat=[0.1, 0.9])

● You now know everything needed to implement a Logistic regression algorithm from
scratch

Implementation

● The LogisticRegression class is written to follow the familiar Scikit-Learn syntax

● The coefficients are set to None at the start - __init__() method
● The fit() method calculates the coefficients
● The predict() method essentially implements the line equation passed through a
sigmoid function
● The _binary_cross_entropy() private function is used to calculate loss at every
iteration:
In [6]:
class LogisticRegression:
'''
A class which implements logistic regression model with gradient descent.
'''
def __init__(self, learning_rate=0.1, n_iterations=1000):
self.learning_rate = learning_rate
self.n_iterations = n_iterations
 self.weights, self.bias = None, None

@staticmethod
def _sigmoid(x):
'''
Private method, used to pass results of the line equation through the sigmoid function.

:param x: float, prediction made by the line equation

:return: float
'''
return 1 / (1 + np.exp(-x))

@staticmethod
def _binary_cross_entropy(y, y_hat):
'''
Private method, used to calculate binary cross entropy value between actual classes
and predicted probabilities.

:param y: array, true class labels

:param y_hat: array, predicted probabilities
:return: float
'''
def safe_log(x):
return 0 if x == 0 else np.log(x)
total = 0
for curr_y, curr_y_hat in zip(y, y_hat):
total += (curr_y * safe_log(curr_y_hat) + (1 - curr_y) * safe_log(1 - curr_y_hat))
return - total / len(y)

def fit(self, X, y):

'''
Used to calculate the coefficient of the logistic regression model.

:param X: array, features

:param y: array, true values
:return: None
'''
# 1. Initialize coefficients
self.weights = np.zeros(X.shape[1])
self.bias = 0

# 2. Perform gradient descent

for i in range(self.n_iterations):
linear_pred = np.dot(X, self.weights) + self.bias
probability = self._sigmoid(linear_pred)
 # Calculate derivatives
partial_w = (1 / X.shape[0]) * (2 * np.dot(X.T, (probability - y)))
partial_d = (1 / X.shape[0]) * (2 * np.sum(probability - y))

# Update the coefficients

self.weights -= self.learning_rate * partial_w
self.bias -= self.learning_rate * partial_d

def predict_proba(self, X):

'''
Calculates prediction probabilities for a given threshold using the line equation
passed through the sigmoid function.

:param X: array, features

:return: array, prediction probabilities
'''
linear_pred = np.dot(X, self.weights) + self.bias
return self._sigmoid(linear_pred)

def predict(self, X, threshold=0.5):

'''
Makes predictions using the line equation passed through the sigmoid function.

:param X: array, features

:param threshold: float, classification threshold
:return: array, predictions
'''
probabilities = self.predict_proba(X)
return [1 if i > threshold else 0 for i in probabilities]

Testing

● We'll use the breast cancer dataset from Scikit-Learn:

In [7]:
from sklearn.datasets import load_breast_cancer

data = load_breast_cancer()
X = data.data
y = data.target

● The below code cell applies train/test split to the dataset:

 In [8]:
from sklearn.model_selection import train_test_split

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)

● You can now initialize and train the model, and afterwards make predictions:
In [9]:
model = LogisticRegression()
model.fit(X_train, y_train)
preds = model.predict(X_test)

● These are the "optimal" weights:

In [10]:
model.weights

● And this is the "optimal" bias:

In [11]:
model.bias

● Let's evaluate the model with two metrics - accuracy score and the confusion matrix
In [12]:
from sklearn.metrics import accuracy_score, confusion_matrix
In [13]:
accuracy_score(y_test, preds)

In [14]:
print(confusion_matrix(y_test, preds))

[[43 0]
[ 6 65]]

● The model seems to be performing good

● Has more false negatives than false positives
Threshold Optimization

● There's no guarantee that 0.5 is the best classification threshold for every
classification problem
● We can change the threshold by altering the threshold parameter of the predict()
method:
● Let's optimize the threshold for accuracy - you can choose any metric you want:
In [15]:
evals = []

for thresh in np.arange(0, 1.01, 0.01):

preds = model.predict(X_test, threshold=thresh)
acc = accuracy_score(y_test, preds)
evals.append({'Threshold': thresh, 'Accuracy': acc})

In [16]:
import pandas as pd
evals_df = pd.DataFrame(evals)

best_thresh = evals_df.sort_values(by='Accuracy', ascending=False).iloc[0]

plt.plot(evals_df['Threshold'], evals_df['Accuracy'], lw=3, c='#087E8B')

plt.scatter(best_thresh['Threshold'], best_thresh['Accuracy'], label=f"Best threshold =
{best_thresh['Threshold']}, Accuracy = {(best_thresh['Accuracy'] * 100):.2f}%", s=250, c='#087E8B')
plt.title('Threshold Optimization', size=20)
plt.xlabel('Threshold', size=14)
plt.ylabel('Accuracy', size=14)
plt.legend()
plt.savefig('images/003_LogisticRegression_threshold_optimization.png', dpi=300,
bbox_inches='tight')
plt.show()

● You can now retrain with the best threshold:

In [17]:
model = LogisticRegression()
model.fit(X_train, y_train)
preds = model.predict(X_test, threshold=0)

● Let's evaluate again for accuracy and confusion matrix:

In [18]:
accuracy_score(y_test, preds)

In [19]:
print(confusion_matrix(y_test, preds))

Comparison with Scikit-Learn

● We want to know if our model is good, so let's compare it with LogisticRegression

model from Scikit-Learn
In [20]:
from sklearn.linear_model import LogisticRegression

lr_model = LogisticRegression()
lr_model.fit(X_train, y_train)
lr_preds = lr_model.predict(X_test)

Increase the number of iterations (max_iter) or scale the data as shown in:
https://scikit-learn.org/stable/modules/preprocessing.html
Please also refer to the documentation for alternative solver options:
https://scikit-learn.org/stable/modules/linear_model.html#logistic-regression
n_iter_i = _check_optimize_result(

In [21]:
accuracy_score(y_test, lr_preds)

In [22]:
print(confusion_matrix(y_test, lr_preds))

● They are more or less the same, accuracy-wise