數據分析-數據處理-pandas處理缺失值

缺失值處理

找出缺失值

# 處理缺失值
import pandas as pd
import numpy as np

# 創建含有缺失值的DateFrame
df = pd.DataFrame(np.random.randn(10,6))
df.iloc[:4, 1] = None
df.iloc[:2, 4:6] = None
df.iloc[6, 3:5] = None
df.iloc[8, 0:2] = None

print(df)

          0         1         2         3         4         5
0  1.905276       NaN -0.978137  1.682227       NaN       NaN
1  0.164089       NaN -1.373336  1.388321       NaN       NaN
2  0.599135       NaN  2.294221  0.737271  0.238057  0.526875
3 -0.481358       NaN -0.519011  0.214852  0.040489  0.599064
4 -0.358184 -0.794557 -0.454329 -1.024130 -0.090786 -0.018543
5 -0.679481 -0.126602  0.726568  1.432684 -1.403756  0.252173
6 -0.842605  1.192915 -0.831709       NaN       NaN -0.073455
7  2.223354 -1.164356 -1.222986  0.009452  1.687198  0.344141
8       NaN       NaN -1.024819  0.689602  1.210335 -0.714473
9 -0.397766  0.739464 -1.535297 -1.868259  0.724042 -1.714549

# 判斷缺失值
# 元素級別的判斷，把對應的所有元素的位置都列出來，元素爲空或者爲NA爲True，否則返回False
res = df.isnull()

res

	0	1	2	3	4	5
0	False	True	False	False	True	True
1	False	True	False	False	True	True
2	False	True	False	False	False	False
3	False	True	False	False	False	False
4	False	False	False	False	False	False
5	False	False	False	False	False	False
6	False	False	False	True	True	False
7	False	False	False	False	False	False
8	True	True	False	False	False	False
9	False	False	False	False	False	False

# 列級別的判斷，只要該該列有空或者NA就返回True，否則返回False
res = df.isnull().any()

res

0     True
1     True
2    False
3     True
4     True
5     True
dtype: bool

# 只顯示存在缺失值的行列，確定缺失值的位置
# drop_duplicates() 去掉重複的行

result = df[df.isnull().values==True].drop_duplicates()
result

	0	1	2	3	4	5
0	1.905276	NaN	-0.978137	1.682227	NaN	NaN
1	0.164089	NaN	-1.373336	1.388321	NaN	NaN
2	0.599135	NaN	2.294221	0.737271	0.238057	0.526875
3	-0.481358	NaN	-0.519011	0.214852	0.040489	0.599064
6	-0.842605	1.192915	-0.831709	NaN	NaN	-0.073455
8	NaN	NaN	-1.024819	0.689602	1.210335	-0.714473

# 獲得爲空或者NA的列索引
res = df.columns[df.isnull().any()==True]
res

Int64Index([0, 1, 3, 4, 5], dtype='int64')

# 獲取每列爲空的數據的個數
num = df.isnull().sum()
num

0    1
1    5
2    0
3    1
4    3
5    2
dtype: int64

# 獲取每行爲空的數據的個數 axis=1 代表行
num = df.isnull().sum(axis=1)
num

0    3
1    3
2    1
3    1
4    0
5    0
6    2
7    0
8    2
9    0
dtype: int64

刪除缺失值所在的行列

# 刪除具有空值的行 不改變原矩陣
df.dropna()

	0	1	2	3	4	5
4	-0.358184	-0.794557	-0.454329	-1.024130	-0.090786	-0.018543
5	-0.679481	-0.126602	0.726568	1.432684	-1.403756	0.252173
7	2.223354	-1.164356	-1.222986	0.009452	1.687198	0.344141
9	-0.397766	0.739464	-1.535297	-1.868259	0.724042	-1.714549

# 刪除具有空值的列
df.dropna(axis=1)

	2
0	-0.978137
1	-1.373336
2	2.294221
3	-0.519011
4	-0.454329
5	0.726568
6	-0.831709
7	-1.222986
8	-1.024819
9	-1.535297

# 所有值爲缺失值才刪除
df.dropna(how='all')

	0	1	2	3	4	5
0	1.905276	NaN	-0.978137	1.682227	NaN	NaN
1	0.164089	NaN	-1.373336	1.388321	NaN	NaN
2	0.599135	NaN	2.294221	0.737271	0.238057	0.526875
3	-0.481358	NaN	-0.519011	0.214852	0.040489	0.599064
4	-0.358184	-0.794557	-0.454329	-1.024130	-0.090786	-0.018543
5	-0.679481	-0.126602	0.726568	1.432684	-1.403756	0.252173
6	-0.842605	1.192915	-0.831709	NaN	NaN	-0.073455
7	2.223354	-1.164356	-1.222986	0.009452	1.687198	0.344141
8	NaN	NaN	-1.024819	0.689602	1.210335	-0.714473
9	-0.397766	0.739464	-1.535297	-1.868259	0.724042	-1.714549

# 至少有四個非空值才保留
df.dropna(thresh=4)

	0	1	2	3	4	5
2	0.599135	NaN	2.294221	0.737271	0.238057	0.526875
3	-0.481358	NaN	-0.519011	0.214852	0.040489	0.599064
4	-0.358184	-0.794557	-0.454329	-1.024130	-0.090786	-0.018543
5	-0.679481	-0.126602	0.726568	1.432684	-1.403756	0.252173
6	-0.842605	1.192915	-0.831709	NaN	NaN	-0.073455
7	2.223354	-1.164356	-1.222986	0.009452	1.687198	0.344141
8	NaN	NaN	-1.024819	0.689602	1.210335	-0.714473
9	-0.397766	0.739464	-1.535297	-1.868259	0.724042	-1.714549

# 刪除這個subset中的含有缺失值的行
df.dropna(subset=[2,4])

	0	1	2	3	4	5
2	0.599135	NaN	2.294221	0.737271	0.238057	0.526875
3	-0.481358	NaN	-0.519011	0.214852	0.040489	0.599064
4	-0.358184	-0.794557	-0.454329	-1.024130	-0.090786	-0.018543
5	-0.679481	-0.126602	0.726568	1.432684	-1.403756	0.252173
7	2.223354	-1.164356	-1.222986	0.009452	1.687198	0.344141
8	NaN	NaN	-1.024819	0.689602	1.210335	-0.714473
9	-0.397766	0.739464	-1.535297	-1.868259	0.724042	-1.714549

缺失值填充

# 缺失值填充0
df.fillna(0)

	0	1	2	3	4	5
0	1.905276	0.000000	-0.978137	1.682227	0.000000	0.000000
1	0.164089	0.000000	-1.373336	1.388321	0.000000	0.000000
2	0.599135	0.000000	2.294221	0.737271	0.238057	0.526875
3	-0.481358	0.000000	-0.519011	0.214852	0.040489	0.599064
4	-0.358184	-0.794557	-0.454329	-1.024130	-0.090786	-0.018543
5	-0.679481	-0.126602	0.726568	1.432684	-1.403756	0.252173
6	-0.842605	1.192915	-0.831709	0.000000	0.000000	-0.073455
7	2.223354	-1.164356	-1.222986	0.009452	1.687198	0.344141
8	0.000000	0.000000	-1.024819	0.689602	1.210335	-0.714473
9	-0.397766	0.739464	-1.535297	-1.868259	0.724042	-1.714549

# 橫向用缺失值前面的值替換缺失值
df.fillna(axis=1, method='ffill')

	0	1	2	3	4	5
0	1.905276	1.905276	-0.978137	1.682227	1.682227	1.682227
1	0.164089	0.164089	-1.373336	1.388321	1.388321	1.388321
2	0.599135	0.599135	2.294221	0.737271	0.238057	0.526875
3	-0.481358	-0.481358	-0.519011	0.214852	0.040489	0.599064
4	-0.358184	-0.794557	-0.454329	-1.024130	-0.090786	-0.018543
5	-0.679481	-0.126602	0.726568	1.432684	-1.403756	0.252173
6	-0.842605	1.192915	-0.831709	-0.831709	-0.831709	-0.073455
7	2.223354	-1.164356	-1.222986	0.009452	1.687198	0.344141
8	NaN	NaN	-1.024819	0.689602	1.210335	-0.714473
9	-0.397766	0.739464	-1.535297	-1.868259	0.724042	-1.714549

# 縱向用缺失值上面的值替換缺失值
df.fillna(axis=0, method='ffill')

	0	1	2	3	4	5
0	1.905276	NaN	-0.978137	1.682227	NaN	NaN
1	0.164089	NaN	-1.373336	1.388321	NaN	NaN
2	0.599135	NaN	2.294221	0.737271	0.238057	0.526875
3	-0.481358	NaN	-0.519011	0.214852	0.040489	0.599064
4	-0.358184	-0.794557	-0.454329	-1.024130	-0.090786	-0.018543
5	-0.679481	-0.126602	0.726568	1.432684	-1.403756	0.252173
6	-0.842605	1.192915	-0.831709	1.432684	-1.403756	-0.073455
7	2.223354	-1.164356	-1.222986	0.009452	1.687198	0.344141
8	2.223354	-1.164356	-1.024819	0.689602	1.210335	-0.714473
9	-0.397766	0.739464	-1.535297	-1.868259	0.724042	-1.714549

# 不同的列用不同的值進行填充
info = {0:0,1:1,2:2,3:3,4:4,5:5}
df.fillna(value=info)

	0	1	2	3	4	5
0	1.905276	1.000000	-0.978137	1.682227	4.000000	5.000000
1	0.164089	1.000000	-1.373336	1.388321	4.000000	5.000000
2	0.599135	1.000000	2.294221	0.737271	0.238057	0.526875
3	-0.481358	1.000000	-0.519011	0.214852	0.040489	0.599064
4	-0.358184	-0.794557	-0.454329	-1.024130	-0.090786	-0.018543
5	-0.679481	-0.126602	0.726568	1.432684	-1.403756	0.252173
6	-0.842605	1.192915	-0.831709	3.000000	4.000000	-0.073455
7	2.223354	-1.164356	-1.222986	0.009452	1.687198	0.344141
8	0.000000	1.000000	-1.024819	0.689602	1.210335	-0.714473
9	-0.397766	0.739464	-1.535297	-1.868259	0.724042	-1.714549

# 對每列出現的替換值有次數限制
df.fillna(value=info, limit=1)

	0	1	2	3	4	5
0	1.905276	1.000000	-0.978137	1.682227	4.000000	5.000000
1	0.164089	NaN	-1.373336	1.388321	NaN	NaN
2	0.599135	NaN	2.294221	0.737271	0.238057	0.526875
3	-0.481358	NaN	-0.519011	0.214852	0.040489	0.599064
4	-0.358184	-0.794557	-0.454329	-1.024130	-0.090786	-0.018543
5	-0.679481	-0.126602	0.726568	1.432684	-1.403756	0.252173
6	-0.842605	1.192915	-0.831709	3.000000	NaN	-0.073455
7	2.223354	-1.164356	-1.222986	0.009452	1.687198	0.344141
8	0.000000	NaN	-1.024819	0.689602	1.210335	-0.714473
9	-0.397766	0.739464	-1.535297	-1.868259	0.724042	-1.714549