在Pandas中，如何计算给定另一列值的一列值的相对概率？

Question

我有两个数据框，vehicles 和 casualties，每个数据框都有一个公共列 Accident_Index:

import pandas as pd

vehicles = pd.DataFrame({'Accident_Index': [1, 1, 2, 3, 3, 4, 4], 
                         'Vehicle_Type': ['car', 'car', 'motorcyle', 'car', 'car', 'car', 'car'],
                         'Sex_Driver': ['male', 'female', 'male', 'female', 'female', 'male', 'male']})

casualties = pd.DataFrame({'Accident_Index': [1, 1, 2, 3, 4],
                           'Casualty_Severity': ['fatal', 'serious', 'fatal', 'light', 'fatal']})

为了便于可视化，这里是vehicles：

   Accident_Index Sex_Driver Vehicle_Type
0               1       male          car
1               1     female          car
2               2       male    motorcyle
3               3     female          car
4               3     female          car
5               4       male          car
6               4       male          car

这里是 casualties:

   Accident_Index Casualty_Severity
0               1             fatal
1               1           serious
2               2             fatal
3               3             light
4               4             fatal

我想计算一下，与涉及女性汽车的事故 drivers 相比，涉及男性汽车 drivers 的事故致死的可能性高多少倍。

到目前为止，我想出了以下解决方案：

dfm = casualties.merge(vehicles, on='Accident_Index')
dfm_cars = dfm.loc[dfm.Vehicle_Type == 'car']

dfm_cars_fatal_male = dfm_cars.isin({'Casualty_Severity': ['fatal'], 'Sex_Driver': ['male']})
male_driver_involved_in_fatal_car_accident = (dfm_cars_fatal_male['Casualty_Severity'] & dfm_cars_fatal_male['Sex_Driver']).sum()

dfm_cars_fatal_female = dfm_cars.isin({'Casualty_Severity': ['fatal'], 'Sex_Driver': ['female']})
female_driver_involved_in_fatal_car_accident = (dfm_cars_fatal_female['Casualty_Severity'] & dfm_cars_fatal_female['Sex_Driver']).sum()

print(male_driver_involved_in_fatal_car_accident / female_driver_involved_in_fatal_car_accident)

在这种情况下，答案是3，因为有两起车祸死亡，一男一女driver，另一男两人drivers.

但是，这段代码似乎不是特别简洁。我该如何重构它？

Answer 1

IIUC，你可以使用 merge + query + groupby:

g = casualties.merge(vehicles, on='Accident_Index')\
        .query("Vehicle_Type == 'car' and Casualty_Severity == 'fatal'")\
        .groupby('Sex_Driver').Sex_Driver.count()

g / g.sum()

Sex_Driver
female    0.25
male      0.75
Name: Sex_Driver, dtype: float64

为简化此操作，您可以使用变量进行查询：

vehicle = 'car'
severity = 'fatal'

然后您可以将 query 步骤重写为：

query("Vehicle_Type == @vehicle and Casualty_Severity == @severity")

这可以更轻松地重用您的代码，例如，如果您想将代码放入函数中并针对各种输入组合进行测试。

在Pandas中，如何计算给定另一列值的一列值的相对概率？

In Pandas, how to calculate the relative probabilities of values of a column given a value of another column?

python

pandas