groupby 在包含字節數組 object 的列上使用 Pandas Dataframe

Question

我有 pandas dataframe 並想對客戶 ID 進行 groupby

 df['rank_col'] = df.groupby('PSEUDO_CUSTOMER_ID')['DB_CREATED_DT'].rank(method='first')

現在問題是 pseudo_customer_ID 看起來像這樣

 [138, 76, 16, 9, 86, 71, 5, 85, 117, 237, 97, 212, 13, 157, 185, 150, 207, 97, 85, 165] 

下面是我對偽客戶 ID 進行價值計數時的快照，

我檢查我得到的單個值低於值

注意：我想對 pseudo_customer_ID 進行 groupby 並按 DB_CREATED_DT 列進行排名

Answer 1

使用bytes function 轉換您的bytearray以允許分組（並獲取可散列類型）：

演示：

df['PSEUDO_CUSTOMER_ID_BYTES'] = df['PSEUDO_CUSTOMER_ID'].apply(bytes)
print(df)

# Output:
                                  PSEUDO_CUSTOMER_ID                           PSEUDO_CUSTOMER_ID_BYTES
0  [138, 76, 16, 9, 86, 71, 5, 85, 117, 237, 97, ...  b'\x8aL\x10\tVG\x05Uu\xeda\xd4\r\x9d\xb9\x96\x...

按PSEUDO_CUSTOMER_ID ：

>>> list(df.groupby('PSEUDO_CUSTOMER_ID'))
...
TypeError: unhashable type: 'bytearray'

按PSEUDO_CUSTOMER_ID_BYTES ：

>>> list(df.groupby('PSEUDO_CUSTOMER_ID_BYTES'))

[(b'\x8aL\x10\tVG\x05Uu\xeda\xd4\r\x9d\xb9\x96\xcfaU\xa5',
                                    PSEUDO_CUSTOMER_ID                           PSEUDO_CUSTOMER_ID_BYTES
  0  [138, 76, 16, 9, 86, 71, 5, 85, 117, 237, 97, ...  b'\x8aL\x10\tVG\x05Uu\xeda\xd4\r\x9d\xb9\x96\x...)]

重要的

如果您確定原始編碼，則可以使用str.decode來獲取str而不是bytes字符串。 這里似乎是latin-1 ：

df['PSEUDO_CUSTOMER_ID_STR'] = df['PSEUDO_CUSTOMER_ID'].decode('latin1'))
print(df.loc[0])

# Output:
PSEUDO_CUSTOMER_ID          [138, 76, 16, 9, 86, 71, 5, 85, 117, 237, 97, ...
PSEUDO_CUSTOMER_ID_BYTES    b'\x8aL\x10\tVG\x05Uu\xeda\xd4\r\x9d\xb9\x96\x...
PSEUDO_CUSTOMER_ID_STR                                 L\tVGUuíaÔ\rÏaU¥
Name: 0, dtype: object

演示：

>>> list(df.groupby('PSEUDO_CUSTOMER_ID_STR'))

[('\x8aL\x10\tVG\x05UuíaÔ\r\x9d¹\x96ÏaU¥',
                                    PSEUDO_CUSTOMER_ID                           PSEUDO_CUSTOMER_ID_BYTES  PSEUDO_CUSTOMER_ID_STR
  0  [138, 76, 16, 9, 86, 71, 5, 85, 117, 237, 97, ...  b'\x8aL\x10\tVG\x05Uu\xeda\xd4\r\x9d\xb9\x96\x...  L\tVGUuíaÔ\rÏaU¥)]