0

我有一个包含大量列的数据框，它们通常遵循以下模式：

'on_calculated_3_things_swell',
'on_calculated_3_things_neap',
'on_calculated_3_things_kts',
'on_calculated_3_things_tov',
'on_calculated_churn_rate_fg2_perc',
'off_calculated_3_things_swell',
'off_calculated_3_things_neap',
'off_calculated_3_things_kts',
'off_calculated_3_things_tov',
'off_calculated_churn_rate_fg2_perc'

它们以 on_ 或 off_ 开头的地方然后具有相同的结尾。我试图从以 off_ 开头但后缀相同的那些中减去以 on_ 开头的那些。它将创建一个以 diff_ 开头的新字段，然后是相同的后缀。这将是一个新的数据框，我想使用带有列表的循环，因为变量的数量会增加。

我试过了：

calc_vars = ['calculated_3_things_swell',
'calculated_3_things_neap',
'calculated_3_things_kts',
'calculated_3_things_tov']

for i in calc_vars:
     df_diff['dif_' + str(i)] = df.['on_' + str(i)] - df.['off_' + str(i)]

但没有这样的运气

4

2 回答 2

0

像这样的东西？

# setup
df = pd.DataFrame.from_records([
 {'string': 'on_calculated_3_things_swell'}, 
 {'string': 'on_calculated_3_things_neap'}, 
 {'string': 'on_calculated_3_things_kts'}, 
 {'string': 'on_calculated_3_things_tov'}, 
 {'string': 'on_calculated_churn_rate_fg2_perc'}, 
 {'string': 'off_calculated_3_things_swell'}, 
 {'string': 'off_calculated_3_things_neap'}, 
 {'string': 'off_calculated_3_things_kts'}, 
 {'string': 'off_calculated_3_things_tov'}, 
 {'string': 'off_calculated_churn_rate_fg2_perc'}])
df['data'] = np.random.rand(len(df))
df

                               string      data
0        on_calculated_3_things_swell  0.047960
1         on_calculated_3_things_neap  0.949035
2          on_calculated_3_things_kts  0.441468
3          on_calculated_3_things_tov  0.144224
4   on_calculated_churn_rate_fg2_perc  0.176003
5       off_calculated_3_things_swell  0.092168
6        off_calculated_3_things_neap  0.300117
7         off_calculated_3_things_kts  0.698156
8         off_calculated_3_things_tov  0.845363
9  off_calculated_churn_rate_fg2_perc  0.384454

# split and subtract
df[['on', 'suffix']] = df['string'].str.split('_', 1, expand=True)
g = df.groupby('on')
diff_series = g.get_group('on').set_index('suffix')['data'].sub(
    g.get_group('off').set_index('suffix')['data']
)
diff_series

suffix
calculated_3_things_swell        -0.044208
calculated_3_things_neap          0.648918
calculated_3_things_kts          -0.256689
calculated_3_things_tov          -0.701139
calculated_churn_rate_fg2_perc   -0.208452
Name: data, dtype: float64

# combine with original df
diff_df = pd.DataFrame({'data': diff_series, 'string': 'dif_' + diff_series.index})
df = pd.concat([df, diff_df], axis=0, join='inner').reset_index(drop=True)
df

                                string      data
0         on_calculated_3_things_swell  0.047960
1          on_calculated_3_things_neap  0.949035
2           on_calculated_3_things_kts  0.441468
3           on_calculated_3_things_tov  0.144224
4    on_calculated_churn_rate_fg2_perc  0.176003
5        off_calculated_3_things_swell  0.092168
6         off_calculated_3_things_neap  0.300117
7          off_calculated_3_things_kts  0.698156
8          off_calculated_3_things_tov  0.845363
9   off_calculated_churn_rate_fg2_perc  0.384454
10       dif_calculated_3_things_swell -0.044208
11        dif_calculated_3_things_neap  0.648918
12         dif_calculated_3_things_kts -0.256689
13         dif_calculated_3_things_tov -0.701139
14  dif_calculated_churn_rate_fg2_perc -0.208452
于 2020-08-12T19:17:36.037 回答
0

删除您之间的点df和['on_'...和['off_'。

此外，请确保它们是受支持的数据类型。如果任何列是字符串类型，它将不起作用，您可以像这样将它们更改为数字

df["column_name"] = pd.to_numeric(df["column_name"])

于 2020-08-12T19:11:10.750 回答