Logistic Regression in Python Case Study – Real-World Classification Project

Logistic Regression in Python – Case Study

Logistic Regression is one of the most widely used machine learning algorithms for solving classification problems. It is simple, fast, and highly interpretable, which makes it a great starting point for beginners in data science and Python machine learning.

In this real-world case study, we will build a complete Logistic Regression project in Python using Scikit-Learn. The goal is to predict whether a customer will purchase a product based on their age and estimated salary.

This type of problem is widely used in marketing analytics, customer targeting, and business decision-making.

By the end of this tutorial, you will clearly understand how to:

• Define a real-world machine learning problem
• Load and explore a dataset
• Prepare and preprocess data
• Train a Logistic Regression model
• Make predictions and probabilities
• Evaluate model performance
• Extract business insights from predictions

---

Business Problem: Customer Purchase Prediction

A company wants to improve its marketing strategy for a new product.

Instead of showing advertisements to all customers, the company wants to target only users who are more likely to buy the product.

Available customer data:

• Age
• Estimated Salary

Target variable:

• Purchased (1 = Yes, 0 = No)

Objective:

Build a Logistic Regression model that predicts whether a customer will purchase the product.

This helps businesses:

• Reduce advertising cost
• Improve conversion rate
• Target the right audience

---

Understanding the Dataset

A sample dataset looks like this:

Age	Estimated Salary	Purchased
19	19000	0
35	20000	0
45	80000	1
50	90000	1
30	30000	0

Features:

• Age
• EstimatedSalary

Target:

• Purchased

---

Import Required Python Libraries

We start by importing the necessary libraries for machine learning:

import pandas as pd
import numpy as np

from sklearn.model_selection import train_test_split
from sklearn.preprocessing import StandardScaler
from sklearn.linear_model import LogisticRegression

from sklearn.metrics import accuracy_score, confusion_matrix, classification_report

---

Load and Inspect the Dataset

data = pd.read_csv("customers.csv")

print(data.head())

Example Output:

   Age  EstimatedSalary  Purchased
0   19            19000          0
1   35            20000          0
2   45            80000          1
3   50            90000          1

---

Exploratory Data Analysis (EDA)

Before training the model, we check data quality:

print(data.info())
print(data.describe())

Why this step is important:

• Detect missing values
• Understand data distribution
• Identify outliers
• Verify data types

---

Define Features and Target Variable

X = data[['Age', 'EstimatedSalary']]
y = data['Purchased']

• X = input features
• y = target output

---

Split Data into Training and Testing Sets

We divide data to evaluate model performance properly:

X_train, X_test, y_train, y_test = train_test_split(
    X,
    y,
    test_size=0.25,
    random_state=42
)

Why splitting is important:

• Prevents overfitting
• Ensures fair evaluation
• Simulates real-world predictions

---

Feature Scaling for Better Performance

Logistic Regression is sensitive to feature scale differences.

scaler = StandardScaler()

X_train = scaler.fit_transform(X_train)
X_test = scaler.transform(X_test)

Example:

Before scaling:

• Age = 45
• Salary = 80000

After scaling:

• Age = 0.85
• Salary = 1.24

---

Train the Logistic Regression Model

model = LogisticRegression()
model.fit(X_train, y_train)

What happens here:

The model learns patterns between:

• Age → Purchase behavior
• Salary → Purchase behavior

---

Make Predictions

predictions = model.predict(X_test)

print(predictions)

Output:

[0 1 1 0 1]

• 0 → Not Purchased
• 1 → Purchased

---

Predict Purchase Probability

One advantage of Logistic Regression is probability output:

probabilities = model.predict_proba(X_test)

print(probabilities[:5])

Example:

[[0.82 0.18]
 [0.21 0.79]
 [0.10 0.90]]

Interpretation:

• Higher second value = higher chance of purchase
• Useful for marketing decision-making

---

Model Evaluation Metrics

1. Accuracy Score

accuracy = accuracy_score(y_test, predictions)
print("Accuracy:", accuracy)

---

2. Confusion Matrix

print(confusion_matrix(y_test, predictions))

Example:

[[60  5]
 [ 6 29]]

Meaning:

• True Negatives: 60
• False Positives: 5
• False Negatives: 6
• True Positives: 29

---

3. Classification Report

print(classification_report(y_test, predictions))

Includes:

• Precision
• Recall
• F1-score

---

Visualization of Customer Data

import matplotlib.pyplot as plt

plt.scatter(
    data['Age'],
    data['EstimatedSalary'],
    c=data['Purchased']
)

plt.xlabel("Age")
plt.ylabel("Estimated Salary")
plt.title("Customer Purchase Behavior")
plt.show()

---

Complete End-to-End Code

import pandas as pd

from sklearn.model_selection import train_test_split
from sklearn.preprocessing import StandardScaler
from sklearn.linear_model import LogisticRegression
from sklearn.metrics import accuracy_score, confusion_matrix

data = pd.read_csv("customers.csv")

X = data[['Age', 'EstimatedSalary']]
y = data['Purchased']

X_train, X_test, y_train, y_test = train_test_split(
    X, y, test_size=0.25, random_state=42
)

scaler = StandardScaler()
X_train = scaler.fit_transform(X_train)
X_test = scaler.transform(X_test)

model = LogisticRegression()
model.fit(X_train, y_train)

predictions = model.predict(X_test)

print("Accuracy:", accuracy_score(y_test, predictions))
print(confusion_matrix(y_test, predictions))

---

Business Insights from the Model

This model helps businesses to:

• Identify potential buyers before advertising
• Reduce unnecessary marketing costs
• Improve targeting strategies
• Increase sales conversion rates
• Understand customer behavior patterns

---

Why Logistic Regression Works Well

Logistic Regression is effective because:

• It is simple and fast
• It works well for binary classification
• It provides probability outputs
• It is easy to interpret
• It performs well on structured data

---

Real-World Applications

This same approach is used in:

• Customer churn prediction
• Credit risk analysis
• Fraud detection systems
• Medical diagnosis models
• Marketing campaign targeting

---

Common Challenges in Real Projects

Real datasets often include:

• Missing values
• Outliers
• Imbalanced classes
• Noisy data
• Irrelevant features

Proper preprocessing improves model accuracy significantly.

---

Conclusion

This case study demonstrated how Logistic Regression can be applied to a real-world customer purchase prediction problem using Python.

By following a complete machine learning pipeline—data loading, preprocessing, feature scaling, training, and evaluation—you can build a reliable classification model for business decision-making.

Logistic Regression remains one of the most important foundational algorithms in machine learning and is widely used in real-world data science applications such as marketing, finance, and healthcare.