Python 3.6.3中含NaN/-NaN的行无法用df.dropna删除，求解决

Ah, I’ve hit this exact snag before—let me break down what’s going on here.

The core issue is that your "nan" and "-nan" values are string literals, not the actual missing value types that pandas recognizes. The dropna() method only targets genuine missing values like:

• numpy.nan
• pandas.NA
• Python's None

When pandas sees a string like "nan", it treats it as a regular non-missing value—so dropna() ignores those rows entirely.

How to Fix This #

First, convert those string "nan"/"-nan" entries to proper missing values, then run dropna():

1. Replace string literals with np.nan
   Use the replace() method to swap the problematic strings:

   import numpy as np
   df = df.replace(['nan', '-nan'], np.nan)
   

   Alternatively, use mask() to target any cell matching those strings:

   df = df.mask(df.isin(['nan', '-nan']))
   

2. Drop the now-recognized missing values
   Now dropna() will work as expected:

   df = df.dropna(axis=0, how="any")
   

How to Verify the Fix #

To confirm the strings are gone and replaced with real NaNs, you can run:

• df.isna().sum(): This will show the count of missing values per column (should now include your former "nan" entries)
• df.applymap(type): Checks the data type of every cell—your converted values should show as float (since np.nan is a float type)

内容的提问来源于stack exchange，提问作者Francesca