如何修改Pandas函数以支持多指标列表的NaN值填充为0

Got it, let's sort this out! The error you're seeing happens because df.isin(Indicators) creates a boolean matrix for your entire DataFrame, but df.loc needs a 1-dimensional boolean array to target the right rows. We just need to narrow the check down to the Indicator column specifically, and also make the function compatible with both single indicator strings and lists of indicators.

Modified Function #

Here's the adjusted version that works for both single values and lists:

import pandas as pd
import numpy as np

def zerofillnaindicator(df, indicators):
    # Handle both single indicator strings and lists
    if not isinstance(indicators, list):
        indicators = [indicators]
    
    # Create a mask targeting rows where Indicator is in our list
    mask = df['Indicator'].isin(indicators)
    
    # Fill NaNs in Value column for those rows with 0
    df.loc[mask, 'Value'] = df.loc[mask, 'Value'].fillna(0)
    return df

How to Test It #

Using your sample DataFrame:

# Sample DataFrame
df = pd.DataFrame({
    'ISO3': ['Australia', 'Austria', 'Belgium', 'Canada', 'Australia', 'Austria', 'Belgium', 'Canada'],
    'Year': [1991]*8,
    'Indicator' : ['Disaster Fatalities']*4 + ['Oil Reserves']*4,
    'Value' : [np.nan, 5, np.nan, 18, np.nan, np.nan, np.nan, np.nan]
})

# Test with a list of indicators
df2 = zerofillnaindicator(df=df, indicators=['Disaster Fatalities', 'Oil Reserves'])
print(df2)

Expected Output #

ISO3  Year             Indicator  Value
0   Australia  1991  Disaster Fatalities    0.0
1     Austria  1991  Disaster Fatalities    5.0
2     Belgium  1991  Disaster Fatalities    0.0
3      Canada  1991  Disaster Fatalities   18.0
4   Australia  1991        Oil Reserves    0.0
5     Austria  1991        Oil Reserves    0.0
6     Belgium  1991        Oil Reserves    0.0
7      Canada  1991        Oil Reserves    0.0

Extra Tip: Avoid Modifying the Original DataFrame #

If you don't want to alter your original DataFrame (which is often a good practice), add a copy step at the start of the function:

def zerofillnaindicator(df, indicators):
    df = df.copy()  # Create a copy to avoid modifying the original
    if not isinstance(indicators, list):
        indicators = [indicators]
    mask = df['Indicator'].isin(indicators)
    df.loc[mask, 'Value'] = df.loc[mask, 'Value'].fillna(0)
    return df

This way, your original df stays untouched, and you get a new DataFrame with the filled values.

内容的提问来源于stack exchange，提问作者Laurens