Pandas - 创建具有相同优先级的类别的分类对象

Question

假设我想订购某些类别。例如，颜色：

Green = Yellow > Red

这里，绿色和黄色的优先级相同，高于红色的优先级。是否有可能创建这样的分类对象？我可以这样做吗？

df['Color'] = pd.Categorical(df['Color'], categories=[('Green', 'Yellow'), 'Red'], ordered=True)

元组('Green', 'Yellow')表示绿色和黄色的优先级相同。

示例输入数据框：

ID    Color
1     Red
2     Yellow
1     Yellow
3     Red
1     Green
2     Red

预期输出是没有重复 ID 的 DataFrame，考虑到颜色的优先级：

ID    Color
1     Yellow
2     Yellow
3     Red

Answer 1

可以解决如下：

由于提供的信息，我们会将黄色和绿色视为相同的优先级。

我们将在这里使用以下方法。

DataFrame.apply 允许我们沿轴应用函数 Docs: https://pandas.pydata.org/pandas-docs/stable/reference/api/pandas.DataFrame.apply.html

sorted 允许我们通过指定键以我们想要的顺序对列表进行排序。文档：https://docs.python.org/3/library/functions.html#sorted 你也可以用 list.sort 做同样的事情，但 sorted 给出了列表的新对象。

import pandas as pd
#create DataFrame
df=pd.DataFrame({'ID':[1,2,1,3,1,2],'Color':['Red','Yellow','Yellow','Red','Green','Red']})

"""
Creating a set_priority list by putting the colors with the highest priority at 
the top and the lowest priority or no priority at the bottom. This list would be 
used as the key in the sorted function below.

Below I am grouping the df with the Id and making a list of all the colors attached to the same ID. Then I am sorting the list based on the priority and choosing the first element from that list as it was asked in the question
"""

set_priority=['Yellow','Green' ,'Red']
result=df.groupby('ID')['Color'].apply(lambda x: sorted(list(x), key=lambda y: set_priority.index(y))[0]).reset_index()

结果

   ID   Color
0   1  Yellow
1   2  Yellow
2   3     Red

Answer 2

使用分类时，可以指定自定义排序顺序。这并没有给出问题中所需的关系，但也许sort_dict 可以用来模拟这样的事情。

import pandas as pd

colors = ["Green", "Red", "Yellow", "Yellow", "Red", "Green"]
df = pd.DataFrame({"Color":colors})
sort_dict = {"Yellow":-1, "Green":1, "Red":6}
df["colorcat"] = pd.Categorical(df['Color'], categories=sorted(sort_dict, key=sort_dict.get), ordered=True)
print(df.sort_values("colorcat"))

    Color colorcat
2  Yellow   Yellow
3  Yellow   Yellow
0   Green    Green
5   Green    Green
1     Red      Red
4     Red      Red

在类别中放置一个元组似乎不起作用。

import pandas as pd

colors = ["Green", "Red", "Yellow", "Yellow", "Red", "Green"]
df = pd.DataFrame({"Color":colors})
df["colorcat"] = pd.Categorical(df['Color'], categories=[("Green", "Yellow"), "Red"], ordered=True)
print(df.sort_values("colorcat"))

    Color colorcat
1     Red      Red
4     Red      Red
0   Green      NaN
2  Yellow      NaN
3  Yellow      NaN
5   Green      NaN

Answer 3

import pandas as pd
# Create an example dataframe
data = {'ID': ['1' , '2', '1', '3', '1', '2'], 
        'Color': ['Red' , 'Yellow' , 'Yellow' , 'Red', 'Green', 'Red']}
df1 = pd.DataFrame(data)

a = df1.join(df1.groupby(['ID'])['Color'].apply(set).rename('m'),
             on=['ID'])['m']

m1 = (a == set({'Green', 'Yellow', 'Red'})) | (a == set({'Green', 'Yellow'}))| (a == set({'Red', 'Yellow'}))
m2 = a == set({'Red'})

m4 = df1['Color'] == 'Yellow'
m5 = df1['Color'] == 'Red'

df1 = df1[(m1 & m4) | (m2 & m5) ]

打印（df1）

    Color ID
1  Yellow  2
2  Yellow  1
3     Red  3