使用 pandas groupby 的动态前 3 名和总百分比
Dynamic top 3 and percentage total using pandas groupby
我有一个如下所示的数据框
id,Name,country,amount,qty
1,ABC,USA,123,4500
1,ABC,USA,156,3210
1,BCE,USA,687,2137
1,DEF,UK,456,1236
1,ABC,nan,216,324
1,DEF,nan,12678,11241
1,nan,nan,637,213
1,BCE,nan,213,543
1,XYZ,KOREA,432,321
1,XYZ,AUS,231,321
sf = pd.read_clipboard(sep=',')
我想做以下事情
a) 根据 amount 为每个 id 和其他选定的列（例如 Name 和 country 获得前 3 名。意思是，我们首先获得基于 id and Name 的前 3 名，然后我们再次获得基于 id and country
的前 3 名
b) 对于每个唯一 ID，找出前 3 项中的每一项对 total amount 的贡献有多大。
所以，我尝试了下面的方法
sf_name = sf.groupby(['id','Name'],dropna=False)['amount'].sum().nlargest(3).reset_index().rename(columns={'amount':'Name_amount'})
sf_country = sf.groupby(['id','country'],dropna=False)['amount'].sum().nlargest(3).reset_index().rename(columns={'amount':'country_amount'})
sf_name['total'] = sf.groupby('id')['amount'].sum()
sf_country['total'] = sf.groupby('id')['amount'].sum()
sf_name['name_pct_total'] = (sf_name['Name_amount']/sf_name['total'])*100
sf_country['country_pct_total'] = (sf_country['country_amount']/sf_country['total'])*100
如您所见，我对每一列重复相同的操作。
但在我的真实数据框中，我必须按 id 进行分组并找到 Top3 并为另外 8 列计算 pct_total %（连同 Name 和 country)
有什么高效、优雅和可扩展的解决方案可以分享吗？
我希望我的输出如下所示
更新 - 完全错误
KeyError Traceback (most recent call last)
C:\Users\Test\AppData\Local\Temp/ipykernel_8720/1850446854.py in <module>
----> 1 df_new.groupby(['unique_key','Resale Customer'],dropna=False)['Revenue Resale EUR'].sum().nlargest(3).reset_index(level=1, name=f'{c}_revenue')
~\Anaconda3\lib\site-packages\pandas\core\series.py in nlargest(self, n, keep)
3834 dtype: int64
3835 """
-> 3836 return algorithms.SelectNSeries(self, n=n, keep=keep).nlargest()
3837
3838 def nsmallest(self, n: int = 5, keep: str = "first") -> Series:
~\Anaconda3\lib\site-packages\pandas\core\algorithms.py in nlargest(self)
1135 @final
1136 def nlargest(self):
-> 1137 return self.compute("nlargest")
1138
1139 @final
~\Anaconda3\lib\site-packages\pandas\core\algorithms.py in compute(self, method)
1181
1182 dropped = self.obj.dropna()
-> 1183 nan_index = self.obj.drop(dropped.index)
1184
1185 if is_extension_array_dtype(dropped.dtype):
~\Anaconda3\lib\site-packages\pandas\util\_decorators.py in wrapper(*args, **kwargs)
309 stacklevel=stacklevel,
310 )
--> 311 return func(*args, **kwargs)
312
313 return wrapper
~\Anaconda3\lib\site-packages\pandas\core\series.py in drop(self, labels, axis, index, columns, level, inplace, errors)
4769 dtype: float64
4770 """
-> 4771 return super().drop(
4772 labels=labels,
4773 axis=axis,
~\Anaconda3\lib\site-packages\pandas\core\generic.py in drop(self, labels, axis, index, columns, level, inplace, errors)
4277 for axis, labels in axes.items():
4278 if labels is not None:
-> 4279 obj = obj._drop_axis(labels, axis, level=level, errors=errors)
4280
4281 if inplace:
~\Anaconda3\lib\site-packages\pandas\core\generic.py in _drop_axis(self, labels, axis, level, errors, consolidate, only_slice)
4321 new_axis = axis.drop(labels, level=level, errors=errors)
4322 else:
-> 4323 new_axis = axis.drop(labels, errors=errors)
4324 indexer = axis.get_indexer(new_axis)
4325
~\Anaconda3\lib\site-packages\pandas\core\indexes\multi.py in drop(self, codes, level, errors)
2234 for level_codes in codes:
2235 try:
-> 2236 loc = self.get_loc(level_codes)
2237 # get_loc returns either an integer, a slice, or a boolean
2238 # mask
~\Anaconda3\lib\site-packages\pandas\core\indexes\multi.py in get_loc(self, key, method)
2880 if keylen == self.nlevels and self.is_unique:
2881 try:
-> 2882 return self._engine.get_loc(key)
2883 except TypeError:
2884 # e.g. test_partial_slicing_with_multiindex partial string slicing
~\Anaconda3\lib\site-packages\pandas\_libs\index.pyx in pandas._libs.index.BaseMultiIndexCodesEngine.get_loc()
~\Anaconda3\lib\site-packages\pandas\_libs\index.pyx in pandas._libs.index.IndexEngine.get_loc()
~\Anaconda3\lib\site-packages\pandas\_libs\index.pyx in pandas._libs.index.IndexEngine.get_loc()
pandas\_libs\hashtable_class_helper.pxi in pandas._libs.hashtable.UInt64HashTable.get_item()
pandas\_libs\hashtable_class_helper.pxi in pandas._libs.hashtable.UInt64HashTable.get_item()
KeyError: 8937472
最简单的是按列表中的列名使用循环，对于 pct_amount 使用 GroupBy.transform 和 sum 每个 id 并划分 amount 列：
dfs = []
cols = ['Name','country']
for c in cols:
df = (sf.groupby(['id',c],dropna=False)['amount'].sum()
.nlargest(3)
.reset_index(level=1, name=f'{c}_amount'))
df[f'{c}_pct_total']=(df[f'{c}_amount'].div(df.groupby('id',dropna=False)[f'{c}_amount']
.transform('sum'))*100)
dfs.append(df)
df = pd.concat(dfs, axis=1)
print (df)
Name Name_amount Name_pct_total country country_amount \
id
1 DEF 13134 89.365177 NaN 13744
1 BCE 900 6.123699 USA 966
1 XYZ 663 4.511125 UK 456
country_pct_total
id
1 90.623764
1 6.369511
1 3.006726
使用 Resale Customer 列名称进行测试::
print (sf)
id Resale Customer country amount qty
0 1 ABC USA 123 4500
1 1 ABC USA 156 3210
2 1 BCE USA 687 2137
3 1 DEF UK 456 1236
4 1 ABC NaN 216 324
5 1 DEF NaN 12678 11241
6 1 NaN NaN 637 213
7 1 BCE NaN 213 543
8 1 XYZ KOREA 432 321
9 1 XYZ AUS 231 321
测试列名称：
print (sf.columns)
Index(['id', 'Resale Customer', 'country', 'amount', 'qty'], dtype='object')
dfs = []
cols = ['Resale Customer','country']
for c in cols:
df = (sf.groupby(['id',c],dropna=False)['amount'].sum()
.nlargest(3)
.reset_index(level=1, name=f'{c}_amount'))
df[f'{c}_pct_total']=(df[f'{c}_amount'].div(df.groupby('id',dropna=False)[f'{c}_amount']
.transform('sum'))*100)
dfs.append(df)
df = pd.concat(dfs, axis=1)
print (df)
Resale Customer Resale Customer_amount Resale Customer_pct_total country \
id
1 DEF 13134 89.365177 NaN
1 BCE 900 6.123699 USA
1 XYZ 663 4.511125 UK
country_amount country_pct_total
id
1 13744 90.623764
1 966 6.369511
1 456 3.006726
可以用 melt 解决，但更复杂：
df = sf.melt(id_vars=['id', 'amount'], value_vars=['Name','country'])
df = (df.groupby(['id','variable', 'value'],dropna=False)['amount']
.sum()
.sort_values(ascending=False)
.groupby(level=[0,1],dropna=False)
.head(3)
.to_frame()
.assign(pct_total=lambda x: x['amount'].div(x.groupby(level=[0,1],dropna=False)['amount'].transform('sum')).mul(100),
g=lambda x: x.groupby(level=[0,1],dropna=False).cumcount())
.set_index('g', append=True)
.reset_index('value')
.unstack(1)
.sort_index(level=1, axis=1)
.droplevel(1)
)
df.columns = df.columns.map(lambda x: f'{x[1]}_{x[0]}')
print (df)
Name_amount Name_pct_total Name_value country_amount country_pct_total \
id
1 13134 89.365177 DEF 13744 90.623764
1 900 6.123699 BCE 966 6.369511
1 663 4.511125 XYZ 456 3.006726
country_value
id
1 NaN
1 USA
1 UK
我有一个如下所示的数据框
id,Name,country,amount,qty
1,ABC,USA,123,4500
1,ABC,USA,156,3210
1,BCE,USA,687,2137
1,DEF,UK,456,1236
1,ABC,nan,216,324
1,DEF,nan,12678,11241
1,nan,nan,637,213
1,BCE,nan,213,543
1,XYZ,KOREA,432,321
1,XYZ,AUS,231,321
sf = pd.read_clipboard(sep=',')
我想做以下事情
a) 根据 amount 为每个 id 和其他选定的列（例如 Name 和 country 获得前 3 名。意思是，我们首先获得基于 id and Name 的前 3 名，然后我们再次获得基于 id and country
b) 对于每个唯一 ID，找出前 3 项中的每一项对 total amount 的贡献有多大。
所以，我尝试了下面的方法
sf_name = sf.groupby(['id','Name'],dropna=False)['amount'].sum().nlargest(3).reset_index().rename(columns={'amount':'Name_amount'})
sf_country = sf.groupby(['id','country'],dropna=False)['amount'].sum().nlargest(3).reset_index().rename(columns={'amount':'country_amount'})
sf_name['total'] = sf.groupby('id')['amount'].sum()
sf_country['total'] = sf.groupby('id')['amount'].sum()
sf_name['name_pct_total'] = (sf_name['Name_amount']/sf_name['total'])*100
sf_country['country_pct_total'] = (sf_country['country_amount']/sf_country['total'])*100
如您所见，我对每一列重复相同的操作。
但在我的真实数据框中，我必须按 id 进行分组并找到 Top3 并为另外 8 列计算 pct_total %（连同 Name 和 country)
有什么高效、优雅和可扩展的解决方案可以分享吗？
我希望我的输出如下所示
更新 - 完全错误
KeyError Traceback (most recent call last)
C:\Users\Test\AppData\Local\Temp/ipykernel_8720/1850446854.py in <module>
----> 1 df_new.groupby(['unique_key','Resale Customer'],dropna=False)['Revenue Resale EUR'].sum().nlargest(3).reset_index(level=1, name=f'{c}_revenue')
~\Anaconda3\lib\site-packages\pandas\core\series.py in nlargest(self, n, keep)
3834 dtype: int64
3835 """
-> 3836 return algorithms.SelectNSeries(self, n=n, keep=keep).nlargest()
3837
3838 def nsmallest(self, n: int = 5, keep: str = "first") -> Series:
~\Anaconda3\lib\site-packages\pandas\core\algorithms.py in nlargest(self)
1135 @final
1136 def nlargest(self):
-> 1137 return self.compute("nlargest")
1138
1139 @final
~\Anaconda3\lib\site-packages\pandas\core\algorithms.py in compute(self, method)
1181
1182 dropped = self.obj.dropna()
-> 1183 nan_index = self.obj.drop(dropped.index)
1184
1185 if is_extension_array_dtype(dropped.dtype):
~\Anaconda3\lib\site-packages\pandas\util\_decorators.py in wrapper(*args, **kwargs)
309 stacklevel=stacklevel,
310 )
--> 311 return func(*args, **kwargs)
312
313 return wrapper
~\Anaconda3\lib\site-packages\pandas\core\series.py in drop(self, labels, axis, index, columns, level, inplace, errors)
4769 dtype: float64
4770 """
-> 4771 return super().drop(
4772 labels=labels,
4773 axis=axis,
~\Anaconda3\lib\site-packages\pandas\core\generic.py in drop(self, labels, axis, index, columns, level, inplace, errors)
4277 for axis, labels in axes.items():
4278 if labels is not None:
-> 4279 obj = obj._drop_axis(labels, axis, level=level, errors=errors)
4280
4281 if inplace:
~\Anaconda3\lib\site-packages\pandas\core\generic.py in _drop_axis(self, labels, axis, level, errors, consolidate, only_slice)
4321 new_axis = axis.drop(labels, level=level, errors=errors)
4322 else:
-> 4323 new_axis = axis.drop(labels, errors=errors)
4324 indexer = axis.get_indexer(new_axis)
4325
~\Anaconda3\lib\site-packages\pandas\core\indexes\multi.py in drop(self, codes, level, errors)
2234 for level_codes in codes:
2235 try:
-> 2236 loc = self.get_loc(level_codes)
2237 # get_loc returns either an integer, a slice, or a boolean
2238 # mask
~\Anaconda3\lib\site-packages\pandas\core\indexes\multi.py in get_loc(self, key, method)
2880 if keylen == self.nlevels and self.is_unique:
2881 try:
-> 2882 return self._engine.get_loc(key)
2883 except TypeError:
2884 # e.g. test_partial_slicing_with_multiindex partial string slicing
~\Anaconda3\lib\site-packages\pandas\_libs\index.pyx in pandas._libs.index.BaseMultiIndexCodesEngine.get_loc()
~\Anaconda3\lib\site-packages\pandas\_libs\index.pyx in pandas._libs.index.IndexEngine.get_loc()
~\Anaconda3\lib\site-packages\pandas\_libs\index.pyx in pandas._libs.index.IndexEngine.get_loc()
pandas\_libs\hashtable_class_helper.pxi in pandas._libs.hashtable.UInt64HashTable.get_item()
pandas\_libs\hashtable_class_helper.pxi in pandas._libs.hashtable.UInt64HashTable.get_item()
KeyError: 8937472
最简单的是按列表中的列名使用循环，对于 pct_amount 使用 GroupBy.transform 和 sum 每个 id 并划分 amount 列：
dfs = []
cols = ['Name','country']
for c in cols:
df = (sf.groupby(['id',c],dropna=False)['amount'].sum()
.nlargest(3)
.reset_index(level=1, name=f'{c}_amount'))
df[f'{c}_pct_total']=(df[f'{c}_amount'].div(df.groupby('id',dropna=False)[f'{c}_amount']
.transform('sum'))*100)
dfs.append(df)
df = pd.concat(dfs, axis=1)
print (df)
Name Name_amount Name_pct_total country country_amount \
id
1 DEF 13134 89.365177 NaN 13744
1 BCE 900 6.123699 USA 966
1 XYZ 663 4.511125 UK 456
country_pct_total
id
1 90.623764
1 6.369511
1 3.006726
使用 Resale Customer 列名称进行测试::
print (sf)
id Resale Customer country amount qty
0 1 ABC USA 123 4500
1 1 ABC USA 156 3210
2 1 BCE USA 687 2137
3 1 DEF UK 456 1236
4 1 ABC NaN 216 324
5 1 DEF NaN 12678 11241
6 1 NaN NaN 637 213
7 1 BCE NaN 213 543
8 1 XYZ KOREA 432 321
9 1 XYZ AUS 231 321
测试列名称：
print (sf.columns)
Index(['id', 'Resale Customer', 'country', 'amount', 'qty'], dtype='object')
dfs = []
cols = ['Resale Customer','country']
for c in cols:
df = (sf.groupby(['id',c],dropna=False)['amount'].sum()
.nlargest(3)
.reset_index(level=1, name=f'{c}_amount'))
df[f'{c}_pct_total']=(df[f'{c}_amount'].div(df.groupby('id',dropna=False)[f'{c}_amount']
.transform('sum'))*100)
dfs.append(df)
df = pd.concat(dfs, axis=1)
print (df)
Resale Customer Resale Customer_amount Resale Customer_pct_total country \
id
1 DEF 13134 89.365177 NaN
1 BCE 900 6.123699 USA
1 XYZ 663 4.511125 UK
country_amount country_pct_total
id
1 13744 90.623764
1 966 6.369511
1 456 3.006726
可以用 melt 解决，但更复杂：
df = sf.melt(id_vars=['id', 'amount'], value_vars=['Name','country'])
df = (df.groupby(['id','variable', 'value'],dropna=False)['amount']
.sum()
.sort_values(ascending=False)
.groupby(level=[0,1],dropna=False)
.head(3)
.to_frame()
.assign(pct_total=lambda x: x['amount'].div(x.groupby(level=[0,1],dropna=False)['amount'].transform('sum')).mul(100),
g=lambda x: x.groupby(level=[0,1],dropna=False).cumcount())
.set_index('g', append=True)
.reset_index('value')
.unstack(1)
.sort_index(level=1, axis=1)
.droplevel(1)
)
df.columns = df.columns.map(lambda x: f'{x[1]}_{x[0]}')
print (df)
Name_amount Name_pct_total Name_value country_amount country_pct_total \
id
1 13134 89.365177 DEF 13744 90.623764
1 900 6.123699 BCE 966 6.369511
1 663 4.511125 XYZ 456 3.006726
country_value
id
1 NaN
1 USA
1 UK