在熊猫中将数据框分组的失败类型方法

Question

I get following error when I try to apply astype(float) method to grouped dataframe in pandas.

ValueError: could not convert string to float:   

Do you know what is the reason why I cannot convert string to float by astype method? and how I can solve this? Below is the code I get error and example data.

def group(self,agg_method):
    df=self.df

    grouped=df.groupby(['Tilt [deg]', 'Azimuth [deg]'],as_index=False)
    groupdf=grouped.agg(agg_method)
    print(groupdf['Azimuth [deg]'][0],len(groupdf['Azimuth [deg]'][0]))
    groupdf['Azimuth [deg]']=groupdf['Azimuth [deg]'].astype(float)  <- I get error here  

Example data

   Tilt [deg] Azimuth [deg]  Glass SHGC  Area of Multiplied Openings [m2]  \
0        90.0        124.48        0.57                           1450.24   
1        90.0         207.3        0.57                            115.66   
2        90.0        207.47        0.57                            115.62   
3        90.0        208.25        0.57                             23.18   
4        90.0        208.26        0.57                            113.12   
5        90.0        214.48        0.57                            451.94   
6        90.0        218.57        0.57                            230.08   
7        90.0        304.46        0.57                             72.66   
8        90.0        304.48        0.57                           1827.53   
9        90.0         34.48        0.57                            917.29 

Answer 1

I believe you need to_numeric with parameter errors='coerce' for convert non numeric to NaN s:

groupdf['Azimuth [deg]']= pd.to_numeric(groupdf['Azimuth [deg]'], errors='coerce')

If need remove all columns which are not parseable and no NaN s values in data is possible use boolean indexing and filter by notnull :

groupdf = groupdf[pd.to_numeric(groupdf['Azimuth [deg]'], errors='coerce').notnull()]

In last version of pandas, 0.21.0 is possible use Series.notna :

groupdf = groupdf[pd.to_numeric(groupdf['Azimuth [deg]'], errors='coerce').notna()]