Python Pandas: How to Use the Pandas get_dummies() Method for One-Hot Encoding

One-hot encoding is a technique for converting categorical data into a numerical format that machine learning algorithms can process. In Pandas, the get_dummies() function handles this conversion automatically, transforming each unique category into its own binary column.

This guide explains how get_dummies() works, covers its key parameters, and demonstrates practical usage patterns including handling missing values and avoiding common pitfalls.

What One-Hot Encoding Does

Categorical variables like "Red", "Blue", and "Green" have no inherent numerical relationship. One-hot encoding converts each category into a separate binary column where 1 indicates the category is present and 0 indicates it is not:

Original	Color_Blue	Color_Green	Color_Red
Red	0	0	1
Blue	1	0	0
Green	0	1	0

This representation allows algorithms that require numerical input to work with categorical data without implying any ordinal relationship between categories.

Syntax and Parameters

pandas.get_dummies(
    data,
    prefix=None,
    prefix_sep='_',
    dummy_na=False,
    columns=None,
    drop_first=False,
    dtype=None
)

Parameter	Description	Default
data	DataFrame, Series, or array-like to encode	Required
prefix	String or list of strings to prepend to new column names	Column name
prefix_sep	Separator between prefix and category value	'_'
dummy_na	If True, adds a column for NaN values	False
columns	List of columns to encode (DataFrame only)	All object/category columns
drop_first	If True, drops the first category to avoid multicollinearity	False
dtype	Data type for the new columns	bool

Encoding a DataFrame

When applied to a DataFrame, get_dummies() automatically detects and encodes all columns with object or category dtype:

import pandas as pd

data = {
    'Color': ['Red', 'Blue', 'Green', 'Blue', 'Red'],
    'Size': ['Small', 'Large', 'Medium', 'Small', 'Large']
}

df = pd.DataFrame(data)
print("Original DataFrame:")
print(df)

df_encoded = pd.get_dummies(df)
print("\nEncoded DataFrame:")
print(df_encoded)

Output:

Original DataFrame:
   Color    Size
0    Red   Small
1   Blue   Large
2  Green  Medium
3   Blue   Small
4    Red   Large

Encoded DataFrame:
   Color_Blue  Color_Green  Color_Red  Size_Large  Size_Medium  Size_Small
0       False        False       True       False        False        True
1        True        False      False        True        False       False
2       False         True      False       False         True       False
3        True        False      False       False        False        True
4       False        False       True        True        False       False

Each unique value in Color and Size becomes its own column. The values are True/False by default.

Getting 0s and 1s Instead of True/False

Most machine learning libraries prefer integer encoding. Set dtype=int to get 0 and 1:

import pandas as pd

data = {
    'Color': ['Red', 'Blue', 'Green', 'Blue', 'Red'],
    'Size': ['Small', 'Large', 'Medium', 'Small', 'Large']
}

df = pd.DataFrame(data)

df_encoded = pd.get_dummies(df, dtype=int)
print(df_encoded)

Output:

   Color_Blue  Color_Green  Color_Red  Size_Large  Size_Medium  Size_Small
0           0            0          1           0            0           1
1           1            0          0           1            0           0
2           0            1          0           0            1           0
3           1            0          0           0            0           1
4           0            0          1           1            0           0

Encoding a Series

get_dummies() also works directly on a Series:

import pandas as pd

data = {
    'Color': ['Red', 'Blue', 'Green', 'Blue', 'Red'],
    'Size': ['Small', 'Large', 'Medium', 'Small', 'Large']
}
df = pd.DataFrame(data)

days = pd.Series(['Monday', 'Tuesday', 'Wednesday', 'Thursday', 'Friday', 'Monday'])
encoded = pd.get_dummies(days, dtype=int)
print(encoded)

Output:

   Friday  Monday  Thursday  Tuesday  Wednesday
0       0       1         0        0          0
1       0       0         0        1          0
2       0       0         0        0          1
3       0       0         1        0          0
4       1       0         0        0          0
5       0       1         0        0          0

Encoding Specific Columns Only

When your DataFrame has a mix of categorical and numeric columns, you may want to encode only certain columns. Use the columns parameter:

import pandas as pd

df = pd.DataFrame({
    'Color': ['Red', 'Blue', 'Green'],
    'Size': ['S', 'M', 'L'],
    'Price': [10.5, 20.0, 15.5]
})

# Encode only the 'Color' column, leave 'Size' and 'Price' unchanged
encoded = pd.get_dummies(df, columns=['Color'], dtype=int)
print(encoded)

Output:

  Size  Price  Color_Blue  Color_Green  Color_Red
0    S   10.5           0            0          1
1    M   20.0           1            0          0
2    L   15.5           0            1          0

note

The Size column remains as-is because it was not included in columns.

Using drop_first to Avoid Multicollinearity

In linear models, having all dummy columns creates multicollinearity - the last column is perfectly predictable from the others. Setting drop_first=True removes the first category from each set of dummies:

import pandas as pd

df = pd.DataFrame({'Color': ['Red', 'Blue', 'Green', 'Red']})

print("Without drop_first:")
print(pd.get_dummies(df, dtype=int))

print("\nWith drop_first=True:")
print(pd.get_dummies(df, drop_first=True, dtype=int))

Output:

Without drop_first:
   Color_Blue  Color_Green  Color_Red
0           0            0          1
1           1            0          0
2           0            1          0
3           0            0          1

With drop_first=True:
   Color_Green  Color_Red
0            0          1
1            0          0
2            1          0
3            0          1

With drop_first=True, Color_Blue is dropped. A row with all zeros in the remaining columns implicitly represents Blue.

tip

Use drop_first=True when building linear regression or logistic regression models to avoid the dummy variable trap. For tree-based models (Random Forest, XGBoost), keeping all columns is generally fine.

Handling Missing Values with dummy_na

By default, NaN values are ignored during encoding. Setting dummy_na=True creates a dedicated column for missing values:

import pandas as pd
import numpy as np

colors = pd.Series(['Red', 'Blue', 'Green', np.nan, 'Red', 'Blue'])

print("Without dummy_na (default):")
print(pd.get_dummies(colors, dtype=int))

print("\nWith dummy_na=True:")
print(pd.get_dummies(colors, dummy_na=True, dtype=int))

Output:

Without dummy_na (default):
   Blue  Green  Red
0     0      0    1
1     1      0    0
2     0      1    0
3     0      0    0
4     0      0    1
5     1      0    0

With dummy_na=True:
   Blue  Green  Red  NaN
0     0      0    1    0
1     1      0    0    0
2     0      1    0    0
3     0      0    0    1
4     0      0    1    0
5     1      0    0    0

Without dummy_na, row 3 has all zeros - there is no way to distinguish it from a potential fourth category. With dummy_na=True, the NaN column explicitly marks missing data.

Customizing Column Name Prefixes

Use the prefix parameter to control the naming of generated columns:

import pandas as pd

df = pd.DataFrame({
    'Color': ['Red', 'Blue'],
    'Size': ['S', 'L']
})

encoded = pd.get_dummies(df, prefix=['c', 's'], prefix_sep='-', dtype=int)
print(encoded)

Output:

   c-Blue  c-Red  s-L  s-S
0       0      1    0    1
1       1      0    1    0

The prefix c replaces Color and s replaces Size, with - as the separator.

Common Mistake: Encoding Numeric Columns Unintentionally

A frequent issue occurs when numeric columns are stored as strings. get_dummies() encodes all object dtype columns, which may include columns that look numeric:

import pandas as pd

df = pd.DataFrame({
    'Category': ['A', 'B', 'A'],
    'ZipCode': ['10001', '10002', '10001'],  # Stored as strings
    'Value': [100, 200, 150]
})

# WRONG: ZipCode gets encoded because it's stored as a string
encoded = pd.get_dummies(df, dtype=int)
print(encoded)

Output:

   Value  Category_A  Category_B  ZipCode_10001  ZipCode_10002
0    100           1           0              1              0
1    200           0           1              0              1
2    150           1           0              1              0

ZipCode was encoded into dummy columns, which is likely unintended.

**The correct approach:**explicitly specify which columns to encode:

import pandas as pd

df = pd.DataFrame({
    'Category': ['A', 'B', 'A'],
    'ZipCode': ['10001', '10002', '10001'],  # Stored as strings
    'Value': [100, 200, 150]
})

# CORRECT: encode only the intended column
encoded = pd.get_dummies(df, columns=['Category'], dtype=int)
print(encoded)

Output:

  ZipCode  Value  Category_A  Category_B
0   10001    100           1           0
1   10002    200           0           1
2   10001    150           1           0

warning

Always use the columns parameter to explicitly specify which columns should be encoded, especially when your DataFrame contains string-typed columns that should not be treated as categories (e.g., IDs, zip codes, names).

Combining Encoded Data with Existing Numeric Columns

A typical workflow keeps numeric columns intact while encoding only categorical ones:

import pandas as pd

df = pd.DataFrame({
    'Name': ['Alice', 'Bob', 'Charlie'],
    'Department': ['Sales', 'Engineering', 'Sales'],
    'Salary': [60000, 85000, 62000],
    'Years': [3, 7, 4]
})

# Encode only 'Department', keep everything else
result = pd.get_dummies(df, columns=['Department'], drop_first=True, dtype=int)
print(result)

Output:

      Name  Salary  Years  Department_Sales
0    Alice   60000      3                 1
1      Bob   85000      7                 0
2  Charlie   62000      4                 1

Quick Reference

Scenario	Parameters
Basic encoding with 0/1	pd.get_dummies(df, dtype=int)
Encode specific columns	pd.get_dummies(df, columns=['col1'], dtype=int)
Avoid multicollinearity	pd.get_dummies(df, drop_first=True, dtype=int)
Handle missing values	pd.get_dummies(df, dummy_na=True, dtype=int)
Custom column names	pd.get_dummies(df, prefix='p', prefix_sep='_', dtype=int)

The get_dummies() function is a quick and effective way to prepare categorical data for analysis and modeling.

For most machine learning pipelines, combining it with columns for targeted encoding and drop_first=True for linear models will cover the majority of use cases.