Python Pandas 使用文本文件创建 Dataframe

Question

我正在尝试使用 Pandas 从原始文本文件创建数据框。该文件包括 3 个类别，类别名称后有与每个类别相关的项目。我能够基于类别创建一个系列，但不知道如何将每个项目类型关联到它们各自的类别并从中创建一个数据框。下面是我的初始代码以及所需的数据帧输出。你能帮我指导一下正确的方法吗？

category = ['Fruits', 'Vegetables', 'Meats']

items='''Fruits
apple
orange
pear
Vegetables
broccoli
squash
carrot
Meats
chicken
beef
lamb'''

Category = pd.Series()

i = 0
for item in items.splitlines():
    if item in category:
        Category = Category.set_value(i, item)
        i += 1
df = pd.DataFrame(Category)
print(df)

所需的数据帧输出：

Category    Item
Fruits      apple
            orange
            pear
Vegetables  broccoli
            squash
            carrot
Meats       chicken
            beef
            lamb

Answer 1

考虑迭代地附加到列表字典而不是系列。然后，将字典转换为数据框。下面的 key 用于输出所需的结果，因为您需要这样的分组的数字：

from io import StringIO
import pandas as pd

txtobj = StringIO('''Fruits
apple
orange
pear
Vegetables
broccoli
squash
carrot
Meats
chicken
beef
lamb''')

items = {'Category':[], 'Item':[]}

for line in txtobj:
    curr_line = line.replace('\n','')
    if curr_line in ['Fruits','Vegetables', 'Meats']:
        curr_category = curr_line       

    if curr_category != curr_line:      
        items['Category'].append(curr_category)
        items['Item'].append(curr_line)

df = pd.DataFrame(items).assign(key=1)
print(df)
#      Category      Item  key
# 0      Fruits     apple    1
# 1      Fruits    orange    1
# 2      Fruits      pear    1
# 3  Vegetables  broccoli    1
# 4  Vegetables    squash    1
# 5  Vegetables    carrot    1
# 6       Meats   chicken    1
# 7       Meats      beef    1
# 8       Meats      lamb    1

print(df['key'].groupby([df['Category'], df['Item']]).count())    
# Category    Item    
# Fruits      apple       1
#             orange      1
#             pear        1
# Meats       beef        1
#             chicken     1
#             lamb        1
# Vegetables  broccoli    1
#             carrot      1
#             squash      1
# Name: key, dtype: int64

Answer 2

这是一个没有 for 循环的解决方案，使用 pandas。

import pandas as pd
category = ['Fruits', 'Vegetables', 'Meats']

items='''Fruits
apple
orange
pear
Vegetables
broccoli
squash
carrot
Meats
chicken
beef
lamb'''

in_df = pd.DataFrame(items.splitlines())

根据该行是否在类别中创建组。

in_df = in_df.assign(group=in_df.isin(category).cumsum())

从每组的第一行创建一个数据框

cat_df = in_df.groupby('group').first()

将每个组的第二行连接回第一行，创建类别水果关系

df_out = in_df.groupby('group').apply(lambda x: x[1:]).reset_index(drop = True).merge(cat_df, left_on='group', right_index=True)

删除分组键并重命名列

df_out = df_out.drop('group',axis=1).rename(columns={'0_x':'Fruit','0_y':'Category'})
print(df_out)

输出：

      Fruit    Category
0     apple      Fruits
1    orange      Fruits
2      pear      Fruits
3  broccoli  Vegetables
4    squash  Vegetables
5    carrot  Vegetables
6   chicken       Meats
7      beef       Meats
8      lamb       Meats

Answer 3

使用：

通过 isin 为检查类别创建掩码
insert new column by where and ffill (fillna 方法 ffill)
通过 boolean indexing and last use reset_index 删除两列中的相同值以获得唯一的单调默认索引。

category = ['Fruits', 'Vegetables', 'Meats']

items='''Fruits
apple
orange
pear
Vegetables
broccoli
squash
carrot
Meats
chicken
beef
lamb'''

df = pd.DataFrame({'Fruit':items.splitlines()})

mask = df['Fruit'].isin(category)
df.insert(0,'Category', df['Fruit'].where(mask).ffill())
df = df[df['Category'] != df['Fruit']].reset_index(drop=True)
print (df)
     Category     Fruit
0      Fruits     apple
1      Fruits    orange
2      Fruits      pear
3  Vegetables  broccoli
4  Vegetables    squash
5  Vegetables    carrot
6       Meats   chicken
7       Meats      beef
8       Meats      lamb

如有必要，最后计数 Categories 和 Fruits 使用 groupby and size:

df1 = df.groupby(['Category','Fruit']).size()
print (df1)
Category    Fruit   
Fruits      apple       1
            orange      1
            pear        1
Meats       beef        1
            chicken     1
            lamb        1
Vegetables  broccoli    1
            carrot      1
            squash      1
dtype: int64

Python Pandas 使用文本文件创建 Dataframe

Python Pandas Create Dataframe using a text file

python

series

dataframe

pandas