How to beginner · 3 min read

How to split features and labels in pandas

Quick answer

Use pandas DataFrame indexing to separate features and labels by selecting columns. Typically, features are all columns except the target label column, which you isolate using df.drop() for features and df['label_column'] for labels.

PREREQUISITES

Python 3.8+
pip install pandas>=1.0
Basic knowledge of pandas DataFrame

Setup

Install pandas if not already installed to handle DataFrame operations.

bash

pip install pandas

Step by step

This example shows how to split features and labels from a pandas DataFrame for PyTorch model training.

python

import pandas as pd

# Sample DataFrame with features and label
 data = {
     'feature1': [5, 6, 7, 8],
     'feature2': [10, 20, 30, 40],
     'label': [0, 1, 0, 1]
 }

df = pd.DataFrame(data)

# Split features (X) and labels (y)
X = df.drop(columns=['label'])  # Features: all columns except 'label'
y = df['label']  # Labels: the 'label' column

print("Features (X):")
print(X)
print("\nLabels (y):")
print(y)

output

Features (X):
   feature1  feature2
0         5        10
1         6        20
2         7        30
3         8        40

Labels (y):
0    0
1    1
2    0
3    1
Name: label, dtype: int64

Common variations

You can split features and labels using column positions or multiple label columns.

python

import pandas as pd

# Using column positions (iloc) to split
X_pos = df.iloc[:, :-1]  # all rows, all columns except last
 y_pos = df.iloc[:, -1]  # all rows, last column

print("Features using iloc:")
print(X_pos)
print("\nLabels using iloc:")
print(y_pos)

# Multiple label columns example
 data_multi_label = {
     'feature1': [1, 2, 3],
     'feature2': [4, 5, 6],
     'label1': [0, 1, 0],
     'label2': [1, 0, 1]
 }

df_multi = pd.DataFrame(data_multi_label)

X_multi = df_multi.drop(columns=['label1', 'label2'])
y_multi = df_multi[['label1', 'label2']]

print("\nFeatures with multiple labels:")
print(X_multi)
print("\nLabels with multiple columns:")
print(y_multi)

output

Features using iloc:
   feature1  feature2
0         5        10
1         6        20
2         7        30
3         8        40

Labels using iloc:
0    0
1    1
2    0
3    1
Name: label, dtype: int64

Features with multiple labels:
   feature1  feature2
0         1        4
1         2        5
2         3        6

Labels with multiple columns:
   label1  label2
0       0       1
1       1       0
2       0       1

Troubleshooting

If you get a KeyError when dropping columns, verify the column names exactly match those in the DataFrame.
Ensure your label column is not included in the features to avoid data leakage during model training.

✅

Key Takeaways

Use df.drop(columns=[...]) to select feature columns excluding labels.
Access label columns directly with df['label_column'] or multiple labels with a list.
Column position indexing with iloc is a flexible alternative for splitting.
Always verify column names to avoid KeyError when dropping columns.
Keep features and labels separate to prepare data correctly for PyTorch training.

Verified 2026-04

Verify ↗