[英]Pandas - Create New Attribute and Value for Groups by Dividing EAV Format Data
Suppose I have a dataframe like such:假设我有一个这样的 dataframe:
import pandas as pd
import numpy as np
data = [[5123, '2021-01-01 00:00:00', 'cash','sales$', 105],
[5123, '2021-01-01 00:00:00', 'cash','items', 20],
[5123, '2021-01-01 00:00:00', 'card','sales$', 355],
[5123, '2021-01-01 00:00:00', 'card','items', 50],
[5123, '2021-01-02 00:00:00', 'cash','sales$', np.nan],
[5123, '2021-01-02 00:00:00', 'cash','items', np.nan],
[5123, '2021-01-02 00:00:00', 'card','sales$', 170],
[5123, '2021-01-02 00:00:00', 'card','items', 35]]
columns = ['Store', 'Date', 'Payment Method', 'Attribute', 'Value']
df = pd.DataFrame(data = data, columns = columns)
Store店铺 | Date日期 | Payment Method付款方法 | Attribute属性 | Value价值 |
---|---|---|---|---|
5123 5123 | 2021-01-01 00:00:00 2021-01-01 00:00:00 | cash现金 | sales$销售额$ | 105 105 |
5123 5123 | 2021-01-01 00:00:00 2021-01-01 00:00:00 | cash现金 | items项目 | 20 20 |
5123 5123 | 2021-01-01 00:00:00 2021-01-01 00:00:00 | card卡片 | sales$销售额$ | 355 355 |
5123 5123 | 2021-01-01 00:00:00 2021-01-01 00:00:00 | card卡片 | items项目 | 50 50 |
5123 5123 | 2021-01-02 00:00:00 2021-01-02 00:00:00 | cash现金 | sales$销售额$ | NaN钠盐 |
5123 5123 | 2021-01-02 00:00:00 2021-01-02 00:00:00 | cash现金 | items项目 | NaN钠盐 |
5123 5123 | 2021-01-02 00:00:00 2021-01-02 00:00:00 | card卡片 | sales$销售额$ | 170 170 |
5123 5123 | 2021-01-02 00:00:00 2021-01-02 00:00:00 | card卡片 | items项目 | 35 35 |
I would like to create a new attribute, called "average item price", which is generated by, for each Store/Date/Payment Method, dividing the sales$ by the items (eg for store 5123, 2021-01-01, cash, I would like to create a new row with an attribute called "average item price", with a value equal to 5.25).我想创建一个名为“平均商品价格”的新属性,它是通过将每个商店/日期/付款方式的销售额除以商品(例如商店 5123、2021-01-01、现金,我想创建一个新行,其中包含一个名为“平均商品价格”的属性,其值等于 5.25)。
I realize that I could pivot this data out, and have one column for sales, one column for items, and divide the two columns, then restack, but is there a better way to do this without having to pivot?我意识到我可以 pivot 这个数据出来,有一列用于销售,一列用于项目,然后将两列分开,然后重新堆叠,但是有没有更好的方法来做到这一点而不必 pivot?
Store店铺 | Date日期 | Payment Method付款方法 | Attribute属性 | Value价值 |
---|---|---|---|---|
5123 5123 | 2021-01-01 00:00:00 2021-01-01 00:00:00 | cash现金 | sales$销售额$ | 105 105 |
5123 5123 | 2021-01-01 00:00:00 2021-01-01 00:00:00 | cash现金 | items项目 | 20 20 |
5123 5123 | 2021-01-01 00:00:00 2021-01-01 00:00:00 | cash现金 | average item price平均商品价格 | 5.25 5.25 |
5123 5123 | 2021-01-01 00:00:00 2021-01-01 00:00:00 | card卡片 | sales$销售额$ | 355 355 |
5123 5123 | 2021-01-01 00:00:00 2021-01-01 00:00:00 | card卡片 | items项目 | 50 50 |
5123 5123 | 2021-01-01 00:00:00 2021-01-01 00:00:00 | card卡片 | average item price平均商品价格 | 7.10 7.10 |
5123 5123 | 2021-01-02 00:00:00 2021-01-02 00:00:00 | cash现金 | sales$销售额$ | NaN钠盐 |
5123 5123 | 2021-01-02 00:00:00 2021-01-02 00:00:00 | cash现金 | items项目 | NaN钠盐 |
5123 5123 | 2021-01-02 00:00:00 2021-01-02 00:00:00 | cash现金 | average item price平均商品价格 | NaN钠盐 |
5123 5123 | 2021-01-02 00:00:00 2021-01-02 00:00:00 | card卡片 | sales$销售额$ | 170 170 |
5123 5123 | 2021-01-02 00:00:00 2021-01-02 00:00:00 | card卡片 | items项目 | 35 35 |
5123 5123 | 2021-01-02 00:00:00 2021-01-02 00:00:00 | card卡片 | average item price平均商品价格 | 4.86 4.86 |
You can use pivot_table
to get the sum of sales/items per group, then compute the average value and merge
with the original data:您可以使用pivot_table
获取每组的销售额/商品的总和,然后计算平均值并与原始数据merge
:
s = (df.pivot_table(index=['Store', 'Date', 'Payment Method'],
columns='Attribute', values='Value', aggfunc='sum')
.assign(avg=lambda d: d['sales$']/d['items'])
['avg']
)
df.merge(s, left_on=['Store', 'Date', 'Payment Method'], right_index=True)
output: output:
Store Date Payment Method Attribute Value avg
0 5123 2021-01-01 00:00:00 cash sales$ 105.0 5.250000
1 5123 2021-01-01 00:00:00 cash items 20.0 5.250000
2 5123 2021-01-01 00:00:00 card sales$ 355.0 7.100000
3 5123 2021-01-01 00:00:00 card items 50.0 7.100000
4 5123 2021-01-02 00:00:00 cash sales$ NaN NaN
5 5123 2021-01-02 00:00:00 cash items NaN NaN
6 5123 2021-01-02 00:00:00 card sales$ 170.0 4.857143
7 5123 2021-01-02 00:00:00 card items 35.0 4.857143
One option is to set the index, do the computation, and use categoricals to get a sorted output that matches yours:一种选择是设置索引,进行计算,并使用分类来获得与您匹配的排序 output:
cols = df.columns[:-1].tolist()
temp = df.set_index(cols)
# computation
summary = temp.xs('sales$', level='Attribute').div(temp.xs('items', level='Attribute'))
# add attribute to index, with a name:
summary = summary.set_index([['average item price'] * len(summary)],
append = True)
summary.index = summary.index.set_names('Attribute', level = -1)
output = pd.concat([temp, summary]).reset_index()
# create categoricals and sort:
dtype = pd.CategoricalDtype(['sales$', 'items', 'average item price'], ordered = True)
output.Attribute = output.Attribute.astype(dtype)
dtype = pd.CategoricalDtype(['cash', 'card'], ordered = True)
output['Payment Method'] = output['Payment Method'].astype(dtype)
output.sort_values(cols)
Store Date Payment Method Attribute Value
0 5123 2021-01-01 00:00:00 cash sales$ 105.000000
1 5123 2021-01-01 00:00:00 cash items 20.000000
8 5123 2021-01-01 00:00:00 cash average item price 5.250000
2 5123 2021-01-01 00:00:00 card sales$ 355.000000
3 5123 2021-01-01 00:00:00 card items 50.000000
9 5123 2021-01-01 00:00:00 card average item price 7.100000
4 5123 2021-01-02 00:00:00 cash sales$ NaN
5 5123 2021-01-02 00:00:00 cash items NaN
10 5123 2021-01-02 00:00:00 cash average item price NaN
6 5123 2021-01-02 00:00:00 card sales$ 170.000000
7 5123 2021-01-02 00:00:00 card items 35.000000
11 5123 2021-01-02 00:00:00 card average item price 4.857143
pivot
and then append
after assigning the "Attribute" as needed:根据需要分配“属性”后, pivot
和append
:
pivoted = df.pivot(["Store", "Date", "Payment Method"], "Attribute", "Value")
output = (df.append(pivoted["sales$"].div(pivoted["items"])
.rename("Value").reset_index()
.assign(Attribute="average item price"), ignore_index=True)
.sort_values(["Store", "Date", "Payment Method"])
.reset_index(drop=True)
)
>>> output
Store Date Payment Method Attribute Value
0 5123 2021-01-01 00:00:00 card sales$ 355.000000
1 5123 2021-01-01 00:00:00 card items 50.000000
2 5123 2021-01-01 00:00:00 card average item price 7.100000
3 5123 2021-01-01 00:00:00 cash sales$ 105.000000
4 5123 2021-01-01 00:00:00 cash items 20.000000
5 5123 2021-01-01 00:00:00 cash average item price 5.250000
6 5123 2021-01-02 00:00:00 card sales$ 170.000000
7 5123 2021-01-02 00:00:00 card items 35.000000
8 5123 2021-01-02 00:00:00 card average item price 4.857143
9 5123 2021-01-02 00:00:00 cash sales$ NaN
10 5123 2021-01-02 00:00:00 cash items NaN
11 5123 2021-01-02 00:00:00 cash average item price NaN
solution 1:方案一:
def function1(dd:pd.DataFrame):
dd1=dd.assign(Value=lambda dd:dd.Value/dd.Value.shift(-1)).head(1).assign(Attribute='average item price')
return pd.concat([dd,dd1])
df.groupby(['Store','Date','Payment Method'],sort=False).apply(function1).reset_index(drop=True).pipe(print)
out:出去:
Store Date Payment Method Attribute Value
0 5123 2021-01-01 00:00:00 cash sales$ 105.000000
1 5123 2021-01-01 00:00:00 cash items 20.000000
2 5123 2021-01-01 00:00:00 cash average item price 5.250000
3 5123 2021-01-01 00:00:00 card sales$ 355.000000
4 5123 2021-01-01 00:00:00 card items 50.000000
5 5123 2021-01-01 00:00:00 card average item price 7.100000
6 5123 2021-01-02 00:00:00 cash sales$ NaN
7 5123 2021-01-02 00:00:00 cash items NaN
8 5123 2021-01-02 00:00:00 cash average item price NaN
9 5123 2021-01-02 00:00:00 card sales$ 170.000000
10 5123 2021-01-02 00:00:00 card items 35.000000
11 5123 2021-01-02 00:00:00 card average item price 4.857143
or you can use pandasql:或者你可以使用pandasql:
def function1(dd:pd.DataFrame):
return dd.sql("""
select * from self union all select tb1.Store,tb1.Date,tb1.[Payment Method],'average item price' Attribute,round(tb1.Value/tb2.value,2) as value from (select * from self where Attribute='sales$') tb1 join (select * from self where Attribute='items') tb2
""")
df.groupby(['Store','Date','Payment Method']).apply(function1).reset_index(drop=True)
声明:本站的技术帖子网页,遵循CC BY-SA 4.0协议,如果您需要转载,请注明本站网址或者原文地址。任何问题请咨询:yoyou2525@163.com.