Os pandas convertem flutuação para int com o valor nulo da nan
#First replace all NaN values with 0 and convert
df['col'] = df['col'].fillna(0).astype(int)
Kwams
#First replace all NaN values with 0 and convert
df['col'] = df['col'].fillna(0).astype(int)math.isnan(n)a[a==0] = np.nanx = float('nan')import math
print math.isnan(float('NaN'))OutputTrue
print math.isnan(1.0)OutputFalse
import pandas as pd
pd.isnull(df.columnName).sum()
pd.notnull(df.columnName).sum()