让Pandas列包含列表,如何将唯一列表元素转移到列？

Question

我写了一个网络刮刀,从产品表中提取信息并构建数据框.数据表有一个Description列,其中包含描述产品的逗号分隔的属性字符串.我想在数据框中为每个唯一属性创建一个列,并使用属性的子字符串填充该列中的行.示例df如下.

PRODUCTS     DATE        DESCRIPTION
Product A    2016-9-12   Steel, Red, High Hardness
Product B    2016-9-11   Blue, Lightweight, Steel
Product C    2016-9-12   Red

我想第一步是将描述拆分成一个列表.

In: df2 = df['DESCRIPTION'].str.split(',')

Out:
DESCRIPTION
['Steel', 'Red', 'High Hardness']
['Blue', 'Lightweight', 'Steel']
['Red']

我想要的输出如下表所示.列名不是特别重要.

PRODUCTS     DATE        STEEL_COL  RED_COL    HIGH HARDNESS_COL  BLUE COL   LIGHTWEIGHT_COL
Product A    2016-9-12   Steel      Red        High Hardness
Product B    2016-9-11   Steel                                    Blue       Lightweight
Product C    2016-9-12              Red

我相信可以使用Pivot设置列,但我不确定建立它们后填充列的最Pythonic方法.任何帮助表示赞赏.

UPDATE

非常感谢您的回答.我选择@ MaxU的回答是正确的,因为它似乎稍微灵活一点,但@ piRSquared得到了一个非常相似的结果,甚至可能被认为是更多的Pythonic方法.我测试了两个版本,都做了我需要的.谢谢!

Answer 1

你可以建立一个稀疏矩阵:

In [27]: df
Out[27]:
    PRODUCTS       DATE                DESCRIPTION
0  Product A  2016-9-12  Steel, Red, High Hardness
1  Product B  2016-9-11   Blue, Lightweight, Steel
2  Product C  2016-9-12                        Red

In [28]: (df.set_index(['PRODUCTS','DATE'])
   ....:    .DESCRIPTION.str.split(',\s*', expand=True)
   ....:    .stack()
   ....:    .reset_index()
   ....:    .pivot_table(index=['PRODUCTS','DATE'], columns=0, fill_value=0, aggfunc='size')
   ....: )
Out[28]:
0                    Blue  High Hardness  Lightweight  Red  Steel
PRODUCTS  DATE
Product A 2016-9-12     0              1            0    1      1
Product B 2016-9-11     1              0            1    0      1
Product C 2016-9-12     0              0            0    1      0

In [29]: (df.set_index(['PRODUCTS','DATE'])
   ....:    .DESCRIPTION.str.split(',\s*', expand=True)
   ....:    .stack()
   ....:    .reset_index()
   ....:    .pivot_table(index=['PRODUCTS','DATE'], columns=0, fill_value='', aggfunc='size')
   ....: )
Out[29]:
0                   Blue High Hardness Lightweight Red Steel
PRODUCTS  DATE
Product A 2016-9-12                  1               1     1
Product B 2016-9-11    1                         1         1
Product C 2016-9-12                                  1

Answer 2

使用 pd.get_dummies

cols = ['PRODUCTS', 'DATE']
pd.get_dummies(
    df.set_index(cols).DESCRIPTION \
      .str.split(',\s*', expand=True).stack()
).groupby(level=cols).sum().astype(int)