Pandas DataFrame 基础操作：创建、计算与合并详解

import pandas as pd
import numpy as np

# 创建DataFrame对象
df = pd.DataFrame([1, 2, 3, 4, 5], columns=['cols'], index=['a','b','c','d','e'])
print df

 cols
a 1
b 2
c 3
d 4
e 5

df2 = pd.DataFrame([[1, 2, 3],[4, 5, 6]], columns=['col1','col2','col3'], index=['a','b'])
print df2

 col1 col2 col3
a 1 2 3
b 4 5 6

df3 = pd.DataFrame(np.array([[1,2],[3,4]]), columns=['col1','col2'], index=['a','b'])
print df3

col1 col2
a 1 2
b 3 4

df4 = pd.DataFrame({'col1':[1,3],'col2':[2,4]},index=['a','b'])
print df4

 col1 col2
a 1 2
b 3 4

# DataFrame对象的基本操作
df2.index

Index([u'a', u'b'], dtype='object')

df2.columns

Index([u'col1', u'col2', u'col3'], dtype='object')

# 根据索引查看数据
df2.loc['a'] 
# 索引为ａ这一行的数据
# df2.iloc[0] 跟上面的操作等价，一个是根据索引名，一个是根据数字索引访问数据

col1 1
col2 2
col3 3
Name: a, dtype: int64

print df2.loc[['a','b']] # 访问多行数据，索引参数为一个列表对象

col1 col2 col3
a 1 2 3
b 4 5 6

print df.loc[df.index[1:3]]

cols
b 2
c 3

# 访问列数据
print df2[['col1','col3']]

col1 col3
a 1 3
b 4 6

# DataFrame元素求和
# 默认是对每列元素求和
print df2.sum()

col1 5
col2 7
col3 9
dtype: int64

# 行求和
print df2.sum(1)

a 6
b 15
dtype: int64

# 对每个元素乘以２
print df2.apply(lambda x:x*2)

col1 col2 col3
a 2 4 6
b 8 10 12

# 对每个元素求平方(支持ndarray一样的向量化操作)
print df2**2

 col1 col2 col3
a 1 4 9
b 16 25 36

列扩充
# 对DataFrame对象进行列扩充
df2['col4'] = ['cnn','rnn']
print df2

 col1 col2 col3 col4
a 1 2 3 cnn
b 4 5 6 rnn

# 也可以通过一个新的DataFrame对象来定义一个新列,索引自动对应
df2['col5'] = pd.DataFrame(['MachineLearning','DeepLearning'],index=['a','b'])
print df2

 col1 col2 col3 col4  col5
a 1 2 3 cnn MachineLearning
b 4 5 6 rnn DeepLearning

# 行进行扩充
print df2.append(pd.DataFrame({'col1':7,'col2':8,'col3':9,'col4':'rcnn','col5':'ReinforcementLearning'},index=['c']))

 col1 col2 col3 col4   col5
a 1 2 3 cnn MachineLearning
b 4 5 6 rnn  DeepLearning
c 7 8 9 rcnn ReinforcementLearning

# 如果在进行　行扩充时候没有，指定ｉｎｄｅｘ的参数，索引会被数字取代
print df2.append({'col1':10,'col2':11,'col3':12,'col4':'frnn','col5':'DRL'},ignore_index=True)

 col1 col2 col3 col4  col5
0 1 2 3 cnn MachineLearning
1 4 5 6 rnn DeepLearning
2 10 11 12 frnn  DRL

# 以上的行扩充，并没有真正修改，df2这个DataFrame对象，除非
df2 = df2.append(pd.DataFrame({'col1':7,'col2':8,'col3':9,'col4':'rcnn','col5':'ReinforcementLearning'},index=['c']))
print df2

 col1 col2 col3 col4   col5
a 1 2 3 cnn MachineLearning
b 4 5 6 rnn  DeepLearning
c 7 8 9 rcnn ReinforcementLearning
c 7 8 9 rcnn ReinforcementLearning

print df2.loc['c']

 col1 col2 col3 col4   col5
c 7 8 9 rcnn ReinforcementLearning
c 7 8 9 rcnn ReinforcementLearning

# DataFrame 对象的合并
df_a = pd.DataFrame(['wang','jing','hui','is','a','master'],columns=['col6'],index=['a','b','c','d','e','f'])
print df_a

 col6
a wang
b jing
c hui
d is
e a
f master

# 默认合并,只保留dfb中的全部索引
dfb = pd.DataFrame([1,2,4,5,6,7],columns=['col1'],index=['a','b','c','d','f','g'])
print dfb.join(df_a)

 col1 col6
a 1 wang
b 2 jing
c 4 hui
d 5 is
f 6 master
g 7 NaN

# 默认合并之接受索引已经存在的值
# 通过指定参数 how，指定合并的方式
print dfb.join(df_a,how='inner') # 合并两个DataFrame对象的交集

 col1 col6
a 1 wang
b 2 jing
c 4 hui
d 5 is
f 6 master

# 合并两个DataFrame对象的并集
print dfb.join(df_a,how='outer')

col1 col6
a 1.0 wang
b 2.0 jing
c 4.0 hui
d 5.0 is
e NaN a
f 6.0 master
g 7.0 NaN