Logistic Regression in Python – Splitting Data for Machine Learning | Train-Test Split Guide

Logistic Regression in Python – Splitting Data

One of the most important steps in machine learning is splitting data into training and testing sets. Before building a Logistic Regression model, we must divide the dataset properly so that we can evaluate how well the model performs on unseen data.

If we train and test a model on the same data, it will give overly optimistic results. That is why data splitting is essential for building reliable and real-world machine learning systems.

In this tutorial, you will learn how to split data for Logistic Regression in Python using Scikit-Learn, along with best practices and real-world understanding.

---

Why Splitting Data is Important

Data splitting helps us measure how well a model generalizes to new data.

It ensures:

• Fair model evaluation
• Prevention of overfitting
• Detection of model performance issues
• Real-world prediction accuracy
• Reliable machine learning workflow

Without splitting, the model may simply memorize the dataset instead of learning patterns.

---

What is Train-Test Split?

Train-test split divides the dataset into two parts:

Training Set

Used to train the Logistic Regression model.

• Usually 70% to 80% of the data
• Model learns patterns here

Testing Set

Used to evaluate the model.

• Usually 20% to 30% of the data
• Model is tested on unseen data

---

Basic Concept

Dataset
   ↓
-------------------------
| Training Data (80%)   |
| Testing Data (20%)    |
-------------------------

---

Import Required Library

We use Scikit-Learn for splitting data.

from sklearn.model_selection import train_test_split

---

Example Dataset

import pandas as pd

data = pd.DataFrame({
    'Age': [22, 25, 30, 35, 40, 45, 50],
    'Salary': [25000, 30000, 45000, 60000, 70000, 85000, 95000],
    'Purchased': [0, 0, 0, 1, 1, 1, 1]
})

print(data)

---

Defining Features and Target

Before splitting, separate input and output variables.

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

• X → Input features
• y → Target variable

---

Performing Train-Test Split

Now we split the dataset.

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

---

Explanation of Parameters

X

Input features dataset.

y

Target labels.

test_size

Percentage of data used for testing.

Example:

test_size = 0.25 → 25% test data, 75% training data

---

random_state

Ensures reproducibility.

random_state = 42 → same split every time

---

Checking Split Results

print("Training set size:", X_train.shape)
print("Testing set size:", X_test.shape)

Example output:

Training set size: (5, 2)
Testing set size: (2, 2)

---

Viewing Training Data

print(X_train)

---

Viewing Testing Data

print(X_test)

---

Why Not Train on Full Data?

If we use all data for training:

• Model memorizes data
• No way to test accuracy
• Overfitting occurs
• Real-world performance becomes unknown

Train-test split prevents this problem.

---

Data Leakage Problem

Data leakage happens when test data influences training.

Example mistakes:

• Scaling before splitting
• Using test data in training
• Feature engineering on full dataset

Correct approach:

Split → Train → Test
NOT
Train + Test together

---

Best Practice Workflow

A standard Logistic Regression pipeline:

1. Load Data
2. Clean Data
3. Feature Engineering
4. Split Data
5. Train Model
6. Evaluate Model

---

Train-Test Split with Scaling (Correct Way)

Scaling should be done AFTER splitting.

from sklearn.preprocessing import StandardScaler

scaler = StandardScaler()

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

Important rule:

• Fit only on training data
• Transform both training and testing data

---

Complete Example Code

import pandas as pd
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import StandardScaler

data = pd.DataFrame({
    'Age': [22, 25, 30, 35, 40, 45, 50],
    'Salary': [25000, 30000, 45000, 60000, 70000, 85000, 95000],
    'Purchased': [0, 0, 0, 1, 1, 1, 1]
})

X = data[['Age', 'Salary']]
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)

print("Data successfully split and prepared")

---

Advantages of Train-Test Split

• Reliable evaluation
• Better generalization
• Prevents overfitting
• Real-world simulation
• Improved model validation

---

Common Mistakes

Avoid:

• Using full dataset for training
• Scaling before splitting
• Ignoring random_state
• Using wrong test size
• Data leakage issues

---

Real-World Importance

Train-test split is used in:

• Fraud detection systems
• Customer prediction models
• Medical diagnosis systems
• Recommendation systems
• Financial risk models

It ensures that machine learning models perform well in real-world scenarios.

---

Conclusion

Splitting data is a critical step in Logistic Regression modeling. It ensures that the model is trained on one portion of data and tested on unseen data, providing a realistic evaluation of performance.

By correctly applying train-test split in Python using Scikit-Learn, you can build more reliable, accurate, and production-ready machine learning models.