Pandas melt()

import pandas as pd

data = {
    'A': [1, 2],
    'B': [4, 5]
}

df = pd.DataFrame(data)

# melt the entire DataFrame
melted_df = pd.melt(df)

print(melted_df)

'''
Output

    variable  value
0        A      1
1        A      2
2        B      4
3        B      5
'''

melt() Syntax

The syntax of the melt() method in Pandas is:

pd.melt(frame, id_vars=None, value_vars=None, var_name=None, value_name='value', col_level=None)

melt() Arguments

The melt() method takes following arguments:

• frame - the DataFrame we want to melt
• id_vars (optional) - a list or a single column name or index to be retained as identifier variables
• value_vars (optional) - a list or a single column name or index indicating which columns to melt
• var_name (optional) - the name to use for the variable column. The default is 'variable'
• value_name (optional) - the name to use for the value column. The default is 'value'
• col_level (optional) - if the input DataFrame has multi-level columns, we can specify the level to melt.

melt() Return Value

The melt() function returns a new DataFrame that represents the melted or reshaped data.

Example1: Reshape DataFrame Using melt()

import pandas as pd

data = {
    'Name': ['Alice', 'Bob', 'Charlie'],
    'Math': [90, 88, 76],
    'Science': [88, 92, 80],
    'History': [78, 85, 90]
}

df = pd.DataFrame(data)

# melt the entire DataFrame 
melted_df = pd.melt(df)

print(melted_df)

Output

      variable   value
0      Name    Alice
1      Name      Bob
2      Name  Charlie
3      Math       90
4      Math       88
5      Math       76
6   Science       88
7   Science       92
8   Science       80
9   History       78
10  History       85
11  History       90

In the above example, we have used the melt() method to melt the entire df DataFrame.

Since we didn't specify any additional arguments, it transforms the DataFrame from its original wide format to a long format.

And melt() also created two columns: variable for the column names and value for the corresponding values.

Example 2: Provide Custom Name to Variable and Value

import pandas as pd

data = {
    'Math': [90, 88, 76],
    'Science': [88, 92, 80],
    'History': [78, 85, 90]
}

df = pd.DataFrame(data)

# melt the entire DataFrame and provide variable and value name
melted_df = pd.melt(df, var_name='Subject', value_name='Score')

print(melted_df)

Output

    Subject  Score
0     Math     90
1     Math     88
2     Math     76
3  Science     88
4  Science     92
5  Science     80
6  History     78
7  History     85
8  History     90

Here,

1. var_name='Subject' names the variable column as Subject.
2. value_name='Score' names the variable column as Score.

Example 3: Preserve Key Information With id_vars in DataFrame Reshaping

import pandas as pd

data = {
    'Name': ['Alice', 'Bob', 'Charlie'],
    'Math': [90, 88, 76],
    'Science': [88, 92, 80],
    'History': [78, 85, 90]
}

df = pd.DataFrame(data)

# melt the DataFrame, keeping 'Name' as an identifier variable
melted_df = pd.melt(df, id_vars=['Name'], var_name='Subject', value_name='Score')

print(melted_df)

Output

     Name    Subject  Score
0    Alice     Math     90
1      Bob     Math     88
2  Charlie     Math     76
3    Alice  Science     88
4      Bob  Science     92
5  Charlie  Science     80
6    Alice  History     78
7      Bob  History     85
8  Charlie  History     90

In this example, we specified id_vars=['Name'], which means we want to keep the Name column as an identifier variable.

As a result, Name is not melted or pivoted, and it appears as a separate column in the melted DataFrame.

Hence, id_vars allows us to specify which columns to preserve in their original form while melting or pivoting the others.

Example 4: Melt Only Specific Columns Using value_vars

import pandas as pd

data = {
    'Name': ['Alice', 'Bob', 'Charlie'],
    'Math': [90, 88, 76],
    'Science': [88, 92, 80],
    'History': [78, 85, 90]
}

df = pd.DataFrame(data)

# melt only specific columns using value_vars
melted_df = pd.melt(df, id_vars=['Name'], value_vars=['Math', 'Science'], var_name='Subject', value_name='Score')

print(melted_df)

Output

     Name    Subject  Score
0    Alice     Math     90
1      Bob     Math     88
2  Charlie     Math     76
3    Alice  Science     88
4      Bob  Science     92
5  Charlie  Science     80

In the above example, we only specified value_vars=['Math', 'Science'], so only the Math and Science columns are melted.

This allows us to control which columns are transformed into the long format while keeping other columns as identifier variables.

Example 5: Melt Multi Level DataFrame

import pandas as pd

# sample DataFrame with multi-level columns
data = {
    ('Jan', 'Sales'): [100, 150, 200],
    ('Feb', 'Sales'): [120, 160, 210],
    ('Jan', 'Profit'): [20, 25, 30],
    ('Feb', 'Profit'): [22, 27, 31],
}

df = pd.DataFrame(data)

# melt the DataFrame, specifying the col_level parameter
melted_df = pd.melt(df, col_level=0, var_name='NewColumn', value_name='Value')

print(melted_df)

Output

        NewColumn  Value
0           Jan    100
1           Jan    150
2           Jan    200
3           Feb    120
4           Feb    160
5           Feb    210
6           Jan     20
7           Jan     25
8           Jan     30
9           Feb     22
10          Feb     27
11          Feb     31

Here, col_level=0 specifies that we want to melt the first level of the column index (i.e., the Jan and Feb columns).

If we set col_level=1, then the second level of the column index would be melted. And the output would have been:

         NewColumn  Value
0           Sales    100
1           Sales    150
2           Sales    200
3           Sales    120
4           Sales    160
5           Sales    210
6          Profit     20
7          Profit     25
8          Profit     30
9          Profit     22
10         Profit     27
11         Profit     31