The integer index or name(s) of the row level to convert into a column level. By default, level=-1, which means that the inner-most row level is converted.

2. fill_valuelink | int or string or dict | optional

The value to fill NaN in the resulting Series/DataFrame. Note that NaN in the original DataFrame will not be filled - only those that appear due to this method will be filled. By default, the NaN is left as is.

Return Value

A Series or a DataFrame.

Examples

Unstacking single-level DataFrames

Consider the following single-level DataFrame:


        
        
            
                
                
                    df = pd.DataFrame({"age":[2,3],"height":[4,5]}, index=["alice","bob"])
df
                
            
                   age  height
alice   2     4
bob     3     5

Calling unstack() on df gives:


        
        
            
                
                
                    df.unstack()
                
            
            age     alice    2
        bob      3
height  alice    4
        bob      5
dtype: int64

Here, note the following:

the return type is Series, with two levels.
the row labels and the column labels in df have merged to form a multi-index.

Unstacking DataFrames with multi-level rows

Consider the following DataFrame with multi-level rows:


        
        
            
                
                
                    index = [("A","alice"), ("A","bob"),("B","cathy"),("B","david")]
multi_index = pd.MultiIndex.from_tuples(index)
df = pd.DataFrame({"age":[2,3,4,5],"height":[6,7,8,9]}, index=multi_index)
df
                
            
                       age   height
A   alice   2      6
    bob     3      7
B   cathy   4      8
    david   5      9

By default, level=-1, which means that the inner-most row level ([alice,bob,cathy,david]) will be converted into a column level:


        
        
            
                
                
                    df.unstack()
                
            
               age                      height
   alice  bob  cathy david  alice  bob  cathy  david
A  2.0    3.0  NaN   NaN    6.0    7.0  NaN    NaN
B  NaN    NaN  4.0   5.0    NaN    NaN  8.0    9.0

Note the following:

the inner-most row level ([alice, bob, cathy, david]) became a column level, and is positioned as the inner-most level.
stacking columns with multi-levels often yield many NaN since, for instance, no data exists about the age of alice in group B.

To specify which levels to convert, pass the level parameter like so:


        
        
            
                
                
                    df.unstack(level=0)
                
            
                    age         height
        A     B     A     B
alice   2.0   NaN   6.0   NaN
bob     3.0   NaN   7.0   NaN
cathy   NaN   4.0   NaN   8.0
david   NaN   5.0   NaN   9.0

Here, level=0 means that the outer-most row level ([A,B]) is converted into a column level.

Specifying fill_value

By default, fill_value=None, which means that NaN in the resulting Series/DataFrame is left as is.

To fill all NaN with a value instead, pass in fill_value like so:


        
        
            
                
                
                    df.unstack(level=0, fill_value="@")
                
            
                    age      height
        A   B    A   B
alice   2   @    6   @
bob     3   @    7   @
cathy   @   4    @   8
david   @   5    @   9

Note that NaN that pre-existed in the original DataFrame will not be filled - only those caused by this unstacking process will be filled by fill_value.

Pandas DataFrame | stack method

Converts the specified column levels to row levels.

chevron_right