根据条件从一列数据框中汇总计数和总值

Question

我有一个数据框，其中包含（比如说）不同产品的发票价格调整值。

例如：

df = pd.DataFrame({'col1': ['A','A','B','B','C','C','A','A','A','A'],
              'sum'  : [10,-10,10,10,0,-10,-10,0,10,0]})

我需要创建一个摘要 table 例如 under for col1 = 'A':

           Count    Value
Positve
Negative
NIL

截至目前，我正在按照以下方式执行此操作：

result = pd.DataFrame(columns=['Count','Value'],index=['Positve','Negative','NIL'])
result.iloc[0,0] = df[(df['col1'] = 'A') & (df['sum'] > 0)]['sum'].count()
result.iloc[0,1] = df[(df['col1'] = 'A') & (df['sum'] > 0)]['sum'].sum()
result.iloc[1,0] = df[(df['col1'] = 'A') & (df['sum'] < 0)]['sum'].count()
result.iloc[1,1] = df[(df['col1'] = 'A') & (df['sum'] < 0)]['sum'].sum()
result.iloc[2,0] = df[(df['col1'] = 'A') & (df['sum'] == 0)]['sum'].count()
result.iloc[2,1] = df[(df['col1'] = 'A') & (df['sum'] == 0)]['sum'].sum()

有没有更好更快的方法来代替为摘要中的每个值编写一行代码table？我在这里想不出任何东西。

Answer 1

首先使用 np.sign with mapping Series.map to new column, then filter only A and aggregate count with sum in GroupBy.agg 命名聚合：

df['new'] = np.sign(df['sum']).map({0:'NIL', 1:'POS', -1:'NEG'})

df1 = df[df['col1'].eq('A')].groupby('new').agg(Count=('new','size'), Val=('sum','sum'))
print (df1)
     Count  Val
new            
NEG      2  -20
NIL      2    0
POS      2   20

如果需要计算所有值，请将另一列添加到 groupby:

df2 = df.groupby(['col1', 'new']).agg(Count=('new','size'), Val=('sum','sum'))
print (df2)
          Count  Val
col1 new            
A    NEG      2  -20
     NIL      2    0
     POS      2   20
B    POS      2   20
C    NEG      1  -10
     NIL      1    0

根据条件从一列数据框中汇总计数和总值

Summarizing count and value totals from a column of dataframe based on conditions

python-3.x

pandas

dataframe

summary