计算 pandas dataframe 中的条目数低于 0

Question

I have a pandas dataframe with many columns of which some are numerical and other categorical.

I want to calculate the number of negative entries in the pandas dataframe. One way is to find which columns are numeric, subset these columns and then use simple syntax to calculate number of entries with negative values, eg (df < 0).sum()

Instead I tried a syntax with apply and lambda function which includes a conditional but I get a message that my syntax is erroneous. Could you please explain to me why and how this idea could be implemented?

data.apply(lambda x: (if (x.dtype == 'int16' or x.dtype == 'float16'): (x<0).sum())).sum()
  File "<ipython-input-75-f329bf4e8cdd>", line 1
    data.apply(lambda x: (if (x.dtype == 'int16' or x.dtype == 'float16'): (x<0).sum())).sum()
                           ^
SyntaxError: invalid syntax

Answer 1

You can use a ternary operator here:

data.apply(lambda x: ).sum()

We thus return 0 (the neutral element of the (ℕ, +, 0) monoid) for non-numerical values.

Note that there are more numerical types than just int16 and float16 , you might want to use np.issubdtype(..., np.number) here:

import numpy as np

data.apply(lambda x: (x < 0).sum() if  else 0).sum()

Answer 2

I think, a simpler solution is:

• use select_dtypes to get a subset of columns of numeric type,
• then use a Numpy count_nonzero function.

As this function counts non-zero values, we have to convert out DataFrame into an array of boolean values, where True values are counted as non-zero.

So to sum up, the whole code can be:

np.count_nonzero(df.select_dtypes(include=np.number) < 0)