JIYIK CN >

Sometimes, we may have a dataset with missing values. There are many ways to replace the missing data using certain methods.

ffill()(Forward Fill) is one of the methods to replace missing values ​​in DataFrame. This method replaces NaN with the previous row or column value.

ffill()Syntax of Pandas Methods

# Python 3.x
dataframe.ffill(axis, inplace, limit, downcast)

ffill()The method takes four optional parameters:

• axisSpecifies where to fill missing values. The value 0 represents rows and 1 represents columns.
• inplaceCan be True or False. True specifies that the changes are made in the current DataFrame, while False creates a separate copy of the new DataFrame with the filled values.
• limitSpecify the maximum number of missing values ​​to be filled consecutively along the axis.
• downcastSpecifies a dictionary of values ​​to be populated for a specific data type.

ffill()Fill missing values ​​in DataFrame using Pandas

Fill missing values ​​along the row axis

In the following code, we have a DataFrame with missing values ​​represented by None or NaN. We have displayed the actual DataFrame and then applied ffill()the method to that DataFrame.

By default, ffill()the method replaces missing values ​​along the row/index axis. NaNs are replaced with the value from the previous row for that cell.

The first row still contains NaNs in the output because there is no preceding row.

Sample code:

# Python 3.x
import pandas as pd

df = pd.DataFrame(
    {
        "C1": [2, 7, None, 4],
        "C2": [None, 2, None, 3],
        "C3": [2, None, 6, 5],
        "C4": [5, 2, 8, None],
    }
)
display(df)
df2 = df.ffill()
display(df2)

Output:

Fill missing values ​​along the column axis

Here, we will specify that axis=1it will fill the missing values ​​by observing the value in the previous column of the corresponding cell.

In the output, all values ​​are filled except two. Since we don't have 1a previous column for column , that value is still NaN.

The value in column 2 is NaN because the corresponding cell in the previous column is also NaN.

Sample code:

# Python 3.x
import pandas as pd

df = pd.DataFrame(
    {
        "C1": [2, 7, None, 4],
        "C2": [None, 2, None, 3],
        "C3": [2, None, 6, 5],
        "C4": [5, 2, 8, None],
    }
)
display(df)
df2 = df.ffill(axis=1)
display(df2)

Output:

Use limitto limit the number of consecutive NaNs to be filled

We can use limitthe parameter to limit the number of consecutive missing values ​​to be filled along the row or column axis.

In the code below, we have the actual DataFrame where the last three rows have consecutive NaNs. limit=2No more than two consecutive NaNs can be filled along the row axis if we specify.

That's why the NaNs in the last row remain unfilled.

Sample code:

# Python 3.x
import pandas as pd

df = pd.DataFrame(
    {
        "C1": [2, 7, None, 4],
        "C2": [4, None, None, None],
        "C3": [6, 6, 6, 5],
        "C4": [None, 2, 8, None],
    }
)
display(df)
df2 = df.ffill(axis=0, limit=2)
display(df2)

Output:

inplaceFill the values ​​in the original DataFrame using

Suppose, instead of copying the DataFrame with filled values ​​in another DataFrame, we want to make changes in the original DataFrame. In this case, we can use inplacethe fill parameter with value True.

Sample code:

# Python 3.x
import pandas as pd

df = pd.DataFrame(
    {
        "C1": [2, 7, None, 4],
        "C2": [4, None, None, None],
        "C3": [6, 6, 6, 5],
        "C4": [None, 2, 8, None],
    }
)
display(df)
df.ffill(inplace=True)
display(df)

Output: