在資料分析和建模的過程中，相當多的時間要用在資料準備上：載入、清理、轉換以及重塑。
這些工作會佔到分析師時間的80%或更更多。
pandas和內建的Python標準庫提供了一組高階的、靈活的、快速的工具，可以讓你輕鬆地將資料規整為想要的格式。

一、處理缺失資料

1.檢測缺失資料

In [10]: string_data = pd.Series(['aardvark', 'artichoke', np.nan, 'avocado'])
In [11]: string_data
Out[11]:
0 aardvark
1 artichoke
2 NaN
3 avocado
dtype: object
In [12]: string_data.isnull()
Out[12]:
0 False
1 False
2 True
3 False
dtype: boo
 

在統計應用中，NA資料可能是不存在的資料或者雖然存在，但是沒有觀察到（例如，資料採
集中發生了了問題）。當進行資料清洗以進行分析時，最好直接對缺失資料進行分析，以判斷數
據採集的問題或缺失資料可能導致的偏差。

Python內建的None值在物件陣列中也可以作為NA：

In [13]: string_data[0] = None
In [14]: string_data.isnull()
Out[14]:
0 True
1 False
2 True
3 False
dtype: bool

二、濾除缺失資料

In [15]: from numpy import nan as NA
In [16]: data = pd.Series([1, NA, 3.5, NA, 7])
In [17]: data.dropna()
Out[17]:
0 1.0
2 3.5
4 7.0
dtype: float64
 

和這個效果一樣

In [18]: data[data.notnull()]
Out[18]:
0 1.0
2 3.5
4 7.0
dtype: float64

DataFrame物件，dropna預設丟棄任何含有缺失值的行

In [19]: data = pd.DataFrame([[1., 6.5, 3.], [1., NA, NA],
....: [NA, NA, NA], [NA, 6.5, 3.]])
In [20]: cleaned = data.dropna()
In [21]: data
Out[21]:
0 1 2
0 1.0 6.5 3.0
1 1.0 NaN NaN
2 NaN NaN NaN
3 NaN 6.5 3.0
In [22]: cleaned
Out[22]:
0 1 2
0 1.0 6.5 3.0
 

傳"how=all’將只丟棄全為NA的那些行

In [23]: data.dropna(how='all')
Out[23]:
0 1 2
0 1.0 6.5 3.0
1 1.0 NaN NaN
3 NaN 6.5 3.0
In [24]: data[4] = NA
In [25]: data
Out[25]:
0 1 2 4
0 1.0 6.5 3.0 NaN
1 1.0 NaN NaN NaN
2 NaN NaN NaN NaN
3 NaN 6.5 3.0 NaN
In [26]: data.dropna(axis=1, how='all')
Out[26]:
0 1 2
0 1.0 6.5 3.0
1 1.0 NaN NaN
2 NaN NaN NaN
3 NaN 6.5 3.0
In [27]: df = pd.DataFrame(np.random.randn(7, 3))
In [28]: df.iloc[:4, 1] = NA
In [29]: df.iloc[:2, 2] = NA
In [30]: df
Out[30]:
0 1 2
0 -0.204708 NaN NaN
1 -0.555730 NaN NaN
2 0.092908 NaN 0.769023
3 1.246435 NaN -1.296221
4 0.274992 0.228913 1.352917
5 0.886429 -2.001637 -0.371843
6 1.669025 -0.438570 -0.539741
In [31]: df.dropna()
Out[31]:
0 1 2
4 0.274992 0.228913 1.352917
5 0.886429 -2.001637 -0.371843
6 1.669025 -0.438570 -0.539741

刪除小於n個非空值的行

In [32]: df.dropna(thresh=2)
Out[32]:
0 1 2
2 0.092908 NaN 0.769023
3 1.246435 NaN -1.296221
4 0.274992 0.228913 1.352917
5 0.886429 -2.001637 -0.371843
6 1.669025 -0.438570 -0.539741

###三、填充缺失資料–fillna()方法

####1.常數呼叫
fillna方法是最主要的函式。通過一個常數呼叫fillna就會將缺失值替換為那個常數值

In [33]: df.fillna(0)
Out[33]:
0 1 2
0 -0.204708 0.000000 0.000000
1 -0.555730 0.000000 0.000000
2 0.092908 0.000000 0.769023
3 1.246435 0.000000 -1.296221
4 0.274992 0.228913 1.352917
5 0.886429 -2.001637 -0.371843
6 1.669025 -0.438570 -0.539741

2.字典呼叫

通過一個字典呼叫fillna，就可以實現對不同的列填充不同的值

In [34]: df.fillna({1: 0.5, 2: 0})
Out[34]:
0 1 2
0 -0.204708 0.500000 0.000000
1 -0.555730 0.500000 0.000000
2 0.092908 0.500000 0.769023
3 1.246435 0.500000 -1.296221
4 0.274992 0.228913 1.352917
5 0.886429 -2.001637 -0.371843
6 1.669025 -0.438570 -0.539741

3.修改物件

fillna預設會返回新物件，但也可以對現有物件進行就地修改
In [35]: _ = df.fillna(0, inplace=True)
In [36]: df
Out[36]:
0 1 2
0 -0.204708 0.000000 0.000000
1 -0.555730 0.000000 0.000000
2 0.092908 0.000000 0.769023
3 1.246435 0.000000 -1.296221
4 0.274992 0.228913 1.352917
5 0.886429 -2.001637 -0.371843
6 1.669025 -0.438570 -0.539741

4.插入數值

對reindexing有效的那些插值方法也可用於fillna

In [37]: df = pd.DataFrame(np.random.randn(6, 3))
In [38]: df.iloc[2:, 1] = NA
In [39]: df.iloc[4:, 2] = NA
In [40]: df
Out[40]:
0 1 2
0 0.476985 3.248944 -1.021228
1 -0.577087 0.124121 0.302614
2 0.523772 NaN 1.343810
3 -0.713544 NaN -2.370232
4 -1.860761 NaN NaN
5 -1.265934 NaN NaN
In [41]: df.fillna(method='ffill')
Out[41]:
0 1 2
0 0.476985 3.248944 -1.021228
1 -0.577087 0.124121 0.302614
2 0.523772 0.124121 1.343810
3 -0.713544 0.124121 -2.370232
4 -1.860761 0.124121 -2.370232
5 -1.265934 0.124121 -2.370232
In [42]: df.fillna(method='ffill', limit=2)
Out[42]:
0 1 2
0 0.476985 3.248944 -1.021228
1 -0.577087 0.124121 0.302614
2 0.523772 0.124121 1.343810
3 -0.713544 0.124121 -2.370232
4 -1.860761 NaN -2.370232
5 -1.265934 NaN -2.370232

5.傳入Series的平均值或中位數

In [43]: data = pd.Series([1., NA, 3.5, NA, 7])
In [44]: data.fillna(data.mean())
Out[44]:
0 1.000000
1 3.833333
2 3.500000
3 3.833333
4 7.000000
dtype: float64

四、資料轉換

1.移除重複資料—drop_duplicated()方法

In [45]: data = pd.DataFrame({'k1': ['one', 'two'] * 3 + ['two'],
....: 'k2': [1, 1, 2, 3, 3, 4, 4]})
In [46]: data
Out[46]:
k1 k2
0 one 1
1 two 1
2 one 2
3 two 3
4 one 3
5 two 4
6 two 4

DataFrame的duplicated方法返回一個布林型Series，表示各行是否是重複行

In [47]: data.duplicated()
Out[47]:
0 False
1 False
2 False
3 False
4 False
5 False
6 True
dtype: bool

drop_duplicates方法，它會返回一個DataFrame，重複的陣列會標為False

In [48]: data.drop_duplicates()
Out[48]:
k1 k2
0 one 1
1 two 1
2 one 2
3 two 3
4 one 3
5 two 4

只希望根據k1列列過濾重複項

In [49]: data['v1'] = range(7)
In [50]: data.drop_duplicates(['k1'])
Out[50]:
k1 k2 v1
0 one 1 0
1 two 1 1

duplicated和drop_duplicates預設保留的是第一個出現的值組合。傳入keep=‘last’則保留最
後一個

In [51]: data.drop_duplicates(['k1', 'k2'], keep='last')
Out[51]:
k1 k2 v1
0 one 1 0
1 two 1 1
2 one 2 2
3 two 3 3
4 one 3 4
6 two 4 6

2.利用函式或對映進行資料轉換

根據陣列、Series或DataFrame列列中的值來實現轉換工作

In [52]: data = pd.DataFrame({'food': ['bacon', 'pulled pork', 'bacon',
....: 'Pastrami', 'corned beef', 'Bacon',
....: 'pastrami', 'honey ham', 'nova lox'],
....: 'ounces': [4, 3, 12, 6, 7.5, 8, 3, 5, 6]})
In [53]: data
Out[53]:
food ounces
0 bacon 4.0
1 pulled pork 3.0
2 bacon 12.0
3 Pastrami 6.0
4 corned beef 7.5
5 Bacon 8.0
6 pastrami 3.0
7 honey ham 5.0
8 nova lox 6.0

新增一列表示該肉類食物來源的動物型別。我們先編寫一個不同肉類到動物的對映

meat_to_animal = {
'bacon': 'pig',
'pulled pork': 'pig',
'pastrami': 'cow',
'corned beef': 'cow',
'honey ham': 'pig',
'nova lox': 'salmon'
}
使⽤用Series的str.lower⽅方法，將各個值轉換為⼩小寫
In [55]: lowercased = data['food'].str.lower()
In [56]: lowercased
Out[56]:
0 bacon
1 pulled pork
2 bacon
3 pastrami
4 corned beef
5 bacon
6 pastrami
7 honey ham
8 nova lox
Name: food, dtype: object
In [57]: data['animal'] = lowercased.map(meat_to_animal)
In [58]: data
Out[58]:
food ounces animal
0 bacon 4.0 pig
1 pulled pork 3.0 pig
2 bacon 12.0 pig
3 Pastrami 6.0 cow
4 corned beef 7.5 cow
5 Bacon 8.0 pig
6 pastrami 3.0 cow
7 honey ham 5.0 pig
8 nova lox 6.0 salmon

也可以傳入一個能夠完成全部這些工作的函式

In [59]: data['food'].map(lambda x: meat_to_animal[x.lower()])
Out[59]:
0 pig
1 pig
2 pig
3 cow
4 cow
5 pig
6 cow
7 pig
8 salmon
Name: food, dtype: object

五、替換值

In [60]: data = pd.Series([1., -999., 2., -999., -1000., 3.])
In [61]: data
Out[61]:
0 1.0
1 -999.0
2 2.0
3 -999.0
4 -1000.0
5 3.0

-999這個值可能是一個表示缺失資料的標記值。要將其替換為pandas能夠理理解的NA值

In [62]: data.replace(-999, np.nan)
Out[62]:
0 1.0
1 NaN
2 2.0
3 NaN
4 -1000.0
5 3.0
dtype: float64

一次性替換多個值

In [63]: data.replace([-999, -1000], np.nan)
Out[63]:
0 1.0
1 NaN
2 2.0
3 NaN
4 NaN
5 3.0
dtype: float64

讓每個值有不不同的替換值，可以傳遞一個替換列列表

In [64]: data.replace([-999, -1000], [np.nan, 0])
Out[64]:
0 1.0
1 NaN
2 2.0
3 NaN
4 0.0
5 3.0
dtype: float64
In [65]: data.replace({-999: np.nan, -1000: 0})
Out[65]:
0 1.0
1 NaN
2 2.0
3 NaN
4 0.0
5 3.0
dtype: float64

六、重新命名軸索引

In [66]: data = pd.DataFrame(np.arange(12).reshape((3, 4)),
....: index=['Ohio', 'Colorado', 'New York'],
....: columns=['one', 'two', 'three', 'four'])
In [67]: transform = lambda x: x[:4].upper()
In [68]: data.index.map(transform)
Out[68]: Index(['OHIO', 'COLO', 'NEW '], dtype='object')
In [69]: data.index = data.index.map(transform)
In [70]: data
Out[70]:
one two three four
OHIO 0 1 2 3
COLO 4 5 6 7
NEW 8 9 10 11

如果想要建立資料集的轉換版（而不是修改原始資料）

In [71]: data.rename(index=str.title, columns=str.upper)
Out[71]:
ONE TWO THREE FOUR
Ohio 0 1 2 3
Colo 4 5 6 7
New 8 9 10 11

rename可以結合字典型物件實現對部分軸標籤的更更新

In [72]: data.rename(index={'OHIO': 'INDIANA'},
....: columns={'three': 'peekaboo'})
Out[72]:
one two peekaboo four
INDIANA 0 1 2 3
COLO 4 5 6 7
NEW 8 9 10 11

就地修改某個資料集，傳入inplace=True即可

In [73]: data.rename(index={'OHIO': 'INDIANA'}, inplace=True)
In [74]: data
Out[74]:
one two three four
INDIANA 0 1 2 3
COLO 4 5 6 7
NEW 8 9 10 11

七、離散化和麵元劃分

為了了便於分析，連續資料常常被離散化或拆分為“面元”（bin）。假設有一組人員資料，而你
希望將它們劃分為不同的年齡組

In [75]: ages = [20, 22, 25, 27, 21, 23, 37, 31, 61, 45, 41, 32]

將這些資料劃分為“18到25”、“26到35”、“35到60”以及“60以上”幾個面元

In [76]: bins = [18, 25, 35, 60, 100]
In [77]: cats = pd.cut(ages, bins)
In [78]: cats
Out[78]:
[(18, 25], (18, 25], (18, 25], (25, 35], (18, 25], ..., (25, 35], (60, 100],
(35,60], (35, 60], (25, 35]]
Length: 12
Categories (4, interval[int64]): [(18, 25] < (25, 35] < (35, 60] < (60, 100]]

pandas返回的是一個特殊的Categorical物件。結果展示了了pandas.cut劃分的⾯面元。你可以將
其看做一組表示面元名稱的字串串

In [79]: cats.codes
Out[79]: array([0, 0, 0, 1, 0, 0, 2, 1, 3, 2, 2, 1], dtype=int8)
In [80]: cats.categories
Out[80]:
IntervalIndex([(18, 25], (25, 35], (35, 60], (60, 100]]
closed='right',
dtype='interval[int64]')

面元計數

In [81]: pd.value_counts(cats)
Out[81]:
(18, 25] 5
(35, 60] 3
(25, 35] 3
(60, 100] 1
dtype: int64

跟“區間”的數學符號一樣，圓括號表示開端，⽽而⽅方括號則表示閉端（包括）。哪邊是閉端可以
通過right=False進行修改

In [82]: pd.cut(ages, [18, 26, 36, 61, 100], right=False)
Out[82]:
[[18, 26), [18, 26), [18, 26), [26, 36), [18, 26), ..., [26, 36), [61, 100),
[36,
61), [36, 61), [26, 36)]
Length: 12
Categories (4, interval[int64]): [[18, 26) < [26, 36) < [36, 61) < [61, 100)]

通過傳遞一個列列表或陣列到labels，設定面元名稱

In [83]: group_names = ['Youth', 'YoungAdult', 'MiddleAged', 'Senior']
In [84]: pd.cut(ages, bins, labels=group_names)
Out[84]:
[Youth, Youth, Youth, YoungAdult, Youth, ..., YoungAdult, Senior, MiddleAged,
Mid
dleAged, YoungAdult]
Length: 12
Categories (4, object): [Youth < YoungAdult < MiddleAged < Senior]

向cut傳入的是面元的數量而不是確切的面元邊界，則它會根據資料的最⼩小值和最大值計算等
長面元。下面這個例例子中，我們將一些均勻分佈的資料分成四組，選項precision=2，限定小
數只有兩位

In [85]: data = np.random.rand(20)
In [86]: pd.cut(data, 4, precision=2)
Out[86]:
[(0.34, 0.55], (0.34, 0.55], (0.76, 0.97], (0.76, 0.97], (0.34, 0.55], ...,
(0.34
, 0.55], (0.34, 0.55], (0.55, 0.76], (0.34, 0.55], (0.12, 0.34]]
Length: 20
Categories (4, interval[float64]): [(0.12, 0.34] < (0.34, 0.55] < (0.55, 0.76]
<
(0.76, 0.97]]

qcut是一個非常類似於cut的函式，它可以根據樣本分位數對資料進行面元劃分。

In [87]: data = np.random.randn(1000) # Normally distributed
In [88]: cats = pd.qcut(data, 4) # Cut into quartiles
In [89]: cats
Out[89]:
[(-0.0265, 0.62], (0.62, 3.928], (-0.68, -0.0265], (0.62, 3.928], (-0.0265,
0.62]
, ..., (-0.68, -0.0265], (-0.68, -0.0265], (-2.95, -0.68], (0.62, 3.928],
(-0.68,
-0.0265]]
Length: 1000
Categories (4, interval[float64]): [(-2.95, -0.68] < (-0.68, -0.0265] <
(-0.0265,
0.62] <
(0.62, 3.928]]
In [90]: pd.value_counts(cats)
Out[90]:
(0.62, 3.928] 250
(-0.0265, 0.62] 250
(-0.68, -0.0265] 250
(-2.95, -0.68] 250
dtype: int64

八、檢測和過濾異常值

In [92]: data = pd.DataFrame(np.random.randn(1000, 4))
In [93]: data.describe()
Out[93]:
0 1 2 3
count 1000.000000 1000.000000 1000.000000 1000.000000
mean 0.049091 0.026112 -0.002544 -0.051827
std 0.996947 1.007458 0.995232 0.998311
min -3.645860 -3.184377 -3.745356 -3.428254
25% -0.599807 -0.612162 -0.687373 -0.747478
50% 0.047101 -0.013609 -0.022158 -0.088274
75% 0.756646 0.695298 0.699046 0.623331
max 2.653656 3.525865 2.735527 3.366626

某列列中絕對值大小超過3的值

In [94]: col = data[2]
In [95]: col[np.abs(col) > 3]
Out[95]:
41 -3.399312
136 -3.745356
Name: 2, dtype: float64

選出全部含有“超過3或－3的值”的行

In [96]: data[(np.abs(data) > 3).any(1)]
Out[96]:
0 1 2 3
41 0.457246 -0.025907 -3.399312 -0.974657
60 1.951312 3.260383 0.963301 1.201206
136 0.508391 -0.196713 -3.745356 -1.520113
235 -0.242459 -3.056990 1.918403 -0.578828
258 0.682841 0.326045 0.425384 -3.428254
322 1.179227 -3.184377 1.369891 -1.074833
544 -3.548824 1.553205 -2.186301 1.277104
635 -0.578093 0.193299 1.397822 3.366626
782 -0.207434 3.525865 0.283070 0.544635
803 -3.645860 0.255475 -0.549574 -1.907459

np.sign(data)可以生成1和-1

In [99]: np.sign(data).head()
Out[99]:
0 1 2 3
0 -1.0 1.0 -1.0 1.0
1 1.0 -1.0 1.0 -1.0
2 1.0 1.0 1.0 -1.0
3 -1.0 -1.0 1.0 -1.0
4 -1.0 1.0 -1.0 -1.0

九、排列列和隨機取樣

利用numpy.random.permutation函式可以輕鬆實現對Series或DataFrame的列列的排列工作
（permuting，隨機重排序）

In [100]: df = pd.DataFrame(np.arange(20).reshape((5, 4)))
In [101]: sampler = np.random.permutation(5)
In [102]: sampler
Out[102]: array([3, 1, 4, 2, 0])

In [103]: df
Out[103]:
0 1 2 3
0 0 1 2 3
1 4 5 6 7
2 8 9 10 11
3 12 13 14 15
4 16 17 18 19
In [104]: df.take(sampler)
Out[104]:
0 1 2 3
3 12 13 14 15
1 4 5 6 7
4 16 17 18 19
2 8 9 10 11
0 0 1 2 3
In [105]: df.sample(n=3)
Out[105]:
0 1 2 3
3 12 13 14 15
4 16 17 18 19
2 8 9 10 11

要通過替換的方式產生樣本（允許重複選擇），可以傳遞replace=True到sample

In [106]: choices = pd.Series([5, 7, -1, 6, 4])
In [107]: draws = choices.sample(n=10, replace=True)
In [108]: draws
Out[108]:
4 4
1 7
4 4
2 -1
0 5
3 6
1 7
4 4
0 5
4 4
dtype: int64

十、計算指標/啞變數

將分類變數量（categorical variable）轉換為“啞變數量”或“指標矩陣”

In [109]: df = pd.DataFrame({'key': ['b', 'b', 'a', 'c', 'a', 'b'],
.....: 'data1': range(6)})
In [110]: pd.get_dummies(df['key'])
Out[110]:
a b c
0 0 1 0
1 0 1 0
2 1 0 0
3 0 0 1
4 1 0 0
5 0 1 0

In [111]: dummies = pd.get_dummies(df['key'], prefix='key')
In [112]: df_with_dummy = df[['data1']].join(dummies)
In [113]: df_with_dummy
Out[113]:
data1 key_a key_b key_c
0 0 0 1 0
1 1 0 1 0
2 2 1 0 0
3 3 0 0 1
4 4 1 0 0
5 5 0 1 0

In [114]: mnames = ['movie_id', 'title', 'genres']
In [115]: movies = pd.read_table('datasets/movielens/movies.dat', sep='::',
.....: header=None, names=mnames, engine='python')
In [116]: movies[:10]
Out[116]:
movie_id title genres
0 1 Toy Story (1995) Animation|Children's|Comedy
1 2 Jumanji (1995) Adventure|Children's|Fantasy
2 3 Grumpier Old Men (1995) Comedy|Romance
3 4 Waiting to Exhale (1995) Comedy|Drama
4 5 Father of the Bride Part II (1995) Comedy
5 6 Heat (1995) Action|Crime|Thriller
6 7 Sabrina (1995) Comedy|Romance
7 8 Tom and Huck (1995) Adventure|Children's
8 9 Sudden Death (1995)
Action
9 10 GoldenEye (1995) Action|Adventure|Thriller

結果去重pd.unique()
方式二：也可以將列表轉化成集合set(list)

In [117]: all_genres = []
In [118]: for x in movies.genres:
.....: all_genres.extend(x.split('|'))
In [119]: genres = pd.unique(all_genres)

構建指標DataFrame的方法之一是從一個全零DataFrame開始

In [121]: zero_matrix = np.zeros((len(movies), len(genres)))
In [122]: dummies = pd.DataFrame(zero_matrix, columns=genres)

十一、pandas的矢量化字串函式

In [167]: data = {'Dave': '[email protected]', 'Steve': '[email protected]',
.....: 'Rob': '[email protected]', 'Wes': np.nan}
In [168]: data = pd.Series(data)
In [169]: data
Out[169]:
Dave [email protected]
Rob [email protected]
Steve [email protected]
Wes NaN
dtype: object
In [170]: data.isnull()
Out[170]:
Dave False
Rob False
Steve False
Wes True
dtype: bool

通過data.map，所有字串串和正則表示式⽅方法都能被應用於（傳入lambda表示式或其他函
數）各個值，但是如果存在NA（null）就會報錯。為了了解決這個問題，Series有一些能夠跳過
NA值的面向陣列方法，進行字串串操作。通過Series的str屬性即可訪問這些方法。例例如，我們可以通過str.contains檢查各個電子郵件地址是否含有"gmail"：

In [171]: data.str.contains('gmail')
Out[171]:
Dave False
Rob True
Steve True
Wes NaN
dtype: object

也可以使用正則表示式，還可以加上任意re選項（如IGNORECASE）

import re
pattern = r'([A-Z0-9._%+-]+)@([A-Z0-9.-]+)\.([A-Z]{2,4})'
In [172]: pattern
Out[172]: '([A-Z0-9._%+-]+)@([A-Z0-9.-]+)\\.([A-Z]{2,4})'
In [173]: data.str.findall(pattern, flags=re.IGNORECASE)
Out[173]:
Dave [(dave, google, com)]
Rob [(rob, gmail, com)]
Steve [(steve, gmail, com)]
Wes NaN
dtype: object
In [174]: matches = data.str.match(pattern, flags=re.IGNORECASE)
In [175]: matches
Out[175]:
Dave True
Rob True
Steve True
Wes NaN
dtype: object

字串進行擷取

In [178]: data.str[:5]
Out[178]:
Dave [email protected]
Rob [email protected]
Steve steve
Wes NaN
dtype: object