Machine Learning - Logistic Regression

Introduction to Logistic Regression

Logistic Regression is a supervised learning algorithm used for classification problems. Unlike Linear Regression which predicts continuous values, Logistic Regression predicts discrete outcomes — often binary (0 or 1).

Think of it like answering a Yes/No question with probability:

• Will the email be spam or not?
• Will the customer buy the product or not?
• Is this tumor malignant or benign?

Real-Life Example: Email Spam Classification

Suppose you're building a model to classify emails as "Spam" (1) or "Not Spam" (0). You collect features like:

• Number of capital words in subject
• Presence of suspicious keywords (e.g., "lottery", "offer")
• Email length

If you apply Linear Regression, it may predict values like 1.2 or -0.4 — but we need a clear binary class! That’s where Logistic Regression shines by converting outputs into probabilities between 0 and 1 using the Sigmoid Function.

The Sigmoid (Logistic) Function

The logistic regression model computes:

y = 1 / (1 + e^(-z))

Where:

• z = b0 + b1*x1 + b2*x2 + ... + bn*xn
• y = predicted probability (0 ≤ y ≤ 1)

We then use a threshold (typically 0.5) to decide:

• If y ≥ 0.5 → class 1 (e.g., Spam)
• If y < 0.5 → class 0 (e.g., Not Spam)

🧠 Question: Why not just use Linear Regression and round the value?

Answer: Linear Regression doesn't output probabilities and is sensitive to outliers. Logistic Regression, through the sigmoid function, maps output between 0 and 1, which is perfect for classification tasks.

Logistic Regression in Python (Step-by-Step)

Problem: Predict if a student will pass an exam

Let’s say you have data of students with their "Hours Studied" and whether they passed (1) or failed (0). We'll build a logistic regression model to predict the outcome.

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
from sklearn.linear_model import LogisticRegression
from sklearn.model_selection import train_test_split
from sklearn.metrics import classification_report, confusion_matrix

# Step 1: Create Dataset
data = {
    "Hours_Studied": [1, 2, 3, 4, 5, 6, 7, 8, 9, 10],
    "Passed":        [0, 0, 0, 0, 1, 1, 1, 1, 1, 1]
}
df = pd.DataFrame(data)

# Step 2: Split features and target
X = df[["Hours_Studied"]]  # input feature
y = df["Passed"]           # target label

# Step 3: Train/Test Split
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)

# Step 4: Train the Logistic Regression model
model = LogisticRegression()
model.fit(X_train, y_train)

# Step 5: Predict on test data
y_pred = model.predict(X_test)

# Step 6: Print performance metrics
print("Confusion Matrix:\n", confusion_matrix(y_test, y_pred))
print("Classification Report:\n", classification_report(y_test, y_pred))

# Step 7: Plot the sigmoid curve
x_range = np.linspace(0, 11, 100)
y_prob = model.predict_proba(x_range.reshape(-1, 1))[:, 1]

plt.scatter(df["Hours_Studied"], df["Passed"], color='red', label='Actual')
plt.plot(x_range, y_prob, color='blue', label='Sigmoid Curve')
plt.xlabel("Hours Studied")
plt.ylabel("Probability of Passing")
plt.title("Logistic Regression - Exam Pass Prediction")
plt.legend()
plt.grid(True)
plt.show()

Code Explanation

• LogisticRegression(): This is the sklearn model for binary classification.
• predict(): Predicts whether a student passed (0 or 1).
• predict_proba(): Gives the probability of the student passing.
• confusion_matrix and classification_report: Help us understand how well the model performs.
• The final plot shows actual data vs. the predicted sigmoid curve.

🧠 Question: What if my data has multiple features like 'Hours Studied', 'Attendance', 'Sleep Hours'?

Answer: Logistic Regression handles multiple features easily. Just use more columns in X = df[[...]]). Scikit-learn automatically handles it internally!

Summary

• Use Logistic Regression for binary classification problems.
• It outputs probabilities using the sigmoid function.
• Apply a threshold to convert probabilities into class predictions.
• Scikit-learn makes it very easy to implement!

What’s Next?

In the next lesson, we’ll explore another popular classification algorithm — K-Nearest Neighbors (KNN).