Python 基于子字符串匹配列
假设您有三个组(A、B、C),它们的组件概述如下:Python 基于子字符串匹配列,python,pandas,join,merge,pattern-matching,Python,Pandas,Join,Merge,Pattern Matching,假设您有三个组(A、B、C),它们的组件概述如下: +-------+-----------+ | Group | Component | +-------+-----------+ | A | 31 | | A | 322 | | A | 323 | | B | 321 | | B | 327 | | B | 33 | | C |
+-------+-----------+
| Group | Component |
+-------+-----------+
| A | 31 |
| A | 322 |
| A | 323 |
| B | 321 |
| B | 327 |
| B | 33 |
| C | 31 |
| C | 32 |
| C | 33 |
+-------+-----------+
这可以存储在名为“组”的数据帧或字典中。请注意,组之间存在重叠
我有以下称为“df”的数据帧(实际表更大):
我想以某种方式将'groups'表与'df'连接起来,这样我就有了另外两列,每一列根据发送方/接收方代码是否包含该组件来显示发送方和接收方关联的组。我只想看到两组都是相同的情况。目标表如下所示:
+--------+----------+--------+--------------+----------------+
| Sender | Receiver | Value | Sender_Group | Receiver_Group |
+--------+----------+--------+--------------+----------------+
| 312345 | 313452 | 1022.1 | A | A |
| 312345 | 313452 | 1022.1 | C | C |
| 320952 | 327901 | 4921.0 | C | C |
+--------+----------+--------+--------------+----------------+
pat = "|".join(groups.Component.astype('str'))
df.insert(0, 'Sender_Group', df['Sender'].str.extract("(" + pat + ')', expand=False))
df.insert(1, 'Receiver_Group', df['Receiver'].str.extract("(" + pat + ')', expand=False))
new_df = df.query('Sender_Group == Receiver_Group')
请注意,尽管327901与组B关联,但320952并非如此,因此未显示。最终的目标是对每组的值进行汇总
我试过这样的方法:
+--------+----------+--------+--------------+----------------+
| Sender | Receiver | Value | Sender_Group | Receiver_Group |
+--------+----------+--------+--------------+----------------+
| 312345 | 313452 | 1022.1 | A | A |
| 312345 | 313452 | 1022.1 | C | C |
| 320952 | 327901 | 4921.0 | C | C |
+--------+----------+--------+--------------+----------------+
pat = "|".join(groups.Component.astype('str'))
df.insert(0, 'Sender_Group', df['Sender'].str.extract("(" + pat + ')', expand=False))
df.insert(1, 'Receiver_Group', df['Receiver'].str.extract("(" + pat + ')', expand=False))
new_df = df.query('Sender_Group == Receiver_Group')
然而,这样做的限制是每个发送方/接收方只能与一个组相关联。我需要一个解决方案,允许他们与多个。有什么想法吗?您可以使用一列来表示发送者(接收者)所属的所有组,作为
列表。然后可以将此列展开为多行,如前所述
给
sender receiver value sender_group
0 312345 313452 1022.1 [A, C]
1 320952 327901 4921.0 [C]
sender receiver value sender_group
0 312345 313452 1022.1 A
1 312345 313452 1022.1 C
2 320952 327901 4921.0 C
对于接收器,程序类似
膨胀
有关展开该列的方法,请参阅。此处给出了一个示例:
indices = np.repeat(df.index.values, df['sender_group'].str.len())
df = df.loc[indices]\
.assign(sender_group=np.concatenate(df['sender_group'].values))\
.reset_index(drop=True)
给
sender receiver value sender_group
0 312345 313452 1022.1 [A, C]
1 320952 327901 4921.0 [C]
sender receiver value sender_group
0 312345 313452 1022.1 A
1 312345 313452 1022.1 C
2 320952 327901 4921.0 C
使用的变量:
groups = pd.DataFrame(
columns=['group', 'component'],
data=[['A', 31],
['A', 322],
['A', 323],
['B', 321],
['B', 327],
['B', 33],
['C', 31],
['C', 32],
['C', 33],]
)
df = pd.DataFrame(
columns=['sender', 'receiver', 'value'],
data=[[312345, 313452, 1022.1],
[320952, 327901, 4921.0]]
)