大熊貓 - pivot_table，同時維持秩序失敗

我有以下的數據幀，其中一週是不是ISO星期，但財政周（1月的第一個星期，六月的最後一週52）：大熊貓 - pivot_table，同時維持秩序失敗

> df 
    domain week count 
0  A 43  5 
1  A 45  1 
2  A 50  1 
3  A 51  4 
4  A  1  3 
5  A  3  12 
6  B 43  1 
7  B 44  1 
8  B 45  4 
9  B 50  11 
10  B  2  3 
11  B  3  12 
12  C 51  6 
13  C  1  14 
14  C  5  1

我想轉動這個表，同時保留一週，以獲得新的數據幀，將看起來像值是計數以下和柱有域：

> new_df 
week A  B  C 
43  5  1 NaN 
44 NaN  1 NaN 
45  1  4 NaN  
50  1 11 NaN 
51  4 NaN  6 
1  3 NaN 14 
2  NaN  3 NaN 
3  12 12 NaN 
5  NaN NaN  1

我嘗試使用追星族和脫膠，但得到此錯誤：

> df = df.groupby(['week'], sort=False)['count'].unstack('domain') 
AttributeError: Cannot access callable attribute 'unstack' of 'SeriesGroupBy' objects, try using the 'apply' method

來源

2017-10-20 Alex J

選項1]可以使用自定義排序weeks指數助手和.loc

In [4810]: weeks = pd.Index(list(range(26, 52)) + list(range(26))) 

In [4819]: dfp = df.groupby(['week','domain'])['count'].sum().unstack() 

In [4820]: dfp.loc[weeks & dfp.index] 
Out[4820]: 
domain  A  B  C 
43  5.0 1.0 NaN 
44  NaN 1.0 NaN 
45  1.0 4.0 NaN 
50  1.0 11.0 NaN 
51  4.0 NaN 6.0 
1  3.0 NaN 14.0 
2  NaN 3.0 NaN 
3  12.0 12.0 NaN 
5  NaN NaN 1.0

選項2]或者，使用pivot

In [4821]: dfp = df.pivot('week', 'domain', 'count') 

In [4822]: dfp.loc[weeks & dfp.index] 
Out[4822]: 
domain  A  B  C 
43  5.0 1.0 NaN 
44  NaN 1.0 NaN 
45  1.0 4.0 NaN 
50  1.0 11.0 NaN 
51  4.0 NaN 6.0 
1  3.0 NaN 14.0 
2  NaN 3.0 NaN 
3  12.0 12.0 NaN 
5  NaN NaN 1.0

選項3]或者，的reindex代替.loc

In [4830]: dfp.reindex(weeks & dfp.index) 
Out[4830]: 
domain  A  B  C 
43  5.0 1.0 NaN 
44  NaN 1.0 NaN 
45  1.0 4.0 NaN 
50  1.0 11.0 NaN 
51  4.0 NaN 6.0 
1  3.0 NaN 14.0 
2  NaN 3.0 NaN 
3  12.0 12.0 NaN 
5  NaN NaN 1.0

詳細

In [4826]: weeks 
Out[4826]: 
Int64Index([26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 
      43, 44, 45, 46, 47, 48, 49, 50, 51, 0, 1, 2, 3, 4, 5, 6, 7, 
      8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 
      25], 
      dtype='int64') 

In [4827]: weeks & dfp.index 
Out[4827]: Int64Index([43, 44, 45, 50, 51, 1, 2, 3, 5], dtype='int64')

來源

2017-10-20 13:02:07 Zero

你需要的week的定製訂單，所以需要ordered categorical與客戶訂單，並省略sort=False：

cats = list(range(26, 52)) + list(range(26)) 
print (cats) 
[26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 
47, 48, 49, 50, 51, 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 
16, 17, 18, 19, 20, 21, 22, 23, 24, 25] 

df['week'] = df['week'].astype('category', ordered=True, categories=cats) 

df = df.groupby(['week','domain'])['count'].sum().unstack() 
print (df) 
domain  A  B  C 
week      
43  5.0 1.0 NaN 
44  NaN 1.0 NaN 
45  1.0 4.0 NaN 
50  1.0 11.0 NaN 
51  4.0 NaN 6.0 
1  3.0 NaN 14.0 
2  NaN 3.0 NaN 
3  12.0 12.0 NaN 
5  NaN NaN 1.0

來源

2017-10-20 11:48:06 jezrael

的問題是，周44和周2是不必要的。第44周應該介於43和45之間，第2周應介於1和3之間。 –

嗯，所以排序是[26,27 ...，51,0,1，...，25]？ – jezrael

大熊貓 - pivot_table，同時維持秩序失敗

回答

相關問題