1、导入
import pandas as pd
import numpy as np
2、数据结构
1、Series
s = pd.Series([1, 2, 3, 4, 5, np.nan, 6, 7])
print(s)
# 0 1.0
# 1 2.0
# 2 3.0
# 3 4.0
# 4 5.0
# 5 NaN
# 6 6.0
# 7 7.0
# dtype: float64
2、DataFrame
dates = pd.date_range('20190101', periods=6)
# index行名,columns列名
df = pd.DataFrame(np.random.randn(6, 4), index=dates, columns=['a', 'b', 'c', 'd'])
print('列选取'.center(50, '-'))
print(df)
# a b c d
# 2019-01-01 -1.294464 0.706790 -0.164825 -0.237432
# 2019-01-02 -1.091822 0.824446 0.748465 -0.191267
# 2019-01-03 -0.755218 1.637604 -1.896371 -0.093815
# 2019-01-04 -2.610031 -0.705783 -1.247235 -1.398978
# 2019-01-05 -0.324550 1.014212 1.375527 -0.409117
# 2019-01-06 -0.512911 0.301417 1.227190 0.771551
# 按照列选择
print('列选取'.center(50, '-'))
print(df['c'])
# 2019-01-01 -0.164825
# 2019-01-02 0.748465
# 2019-01-03 -1.896371
# 2019-01-04 -1.247235
# 2019-01-05 1.375527
# 2019-01-06 1.227190
# Freq: D, Name: c, dtype: float64
3、创建特定数据的DataFrame
df_1 = pd.DataFrame({
'A': 1, # 直接赋值,如果index数量大于1,默认此列按照这个值重复
'B': pd.Timestamp('20190930'), # 直接赋值Timestamp格式,多行,重复这个值
'C': pd.Series(2, index=list(range(4)), dtype='float'), # Series结构使用index,定义出整个函数的值
'D': pd.Categorical([1, 2, 3, 4]),
'E': ['a', 'b', 'c', 'd'],
'F': 'beer',
'G': [1, 5, 4, 4]
})
print(df_1)
# A B C D E F G
# 0 1 2019-09-30 2.0 1 a beer 1
# 1 1 2019-09-30 2.0 2 b beer 5
# 2 1 2019-09-30 2.0 3 c beer 4
# 3 1 2019-09-30 2.0 4 d beer 4
4、DataFrame常用属性及排序
print('types'.center(50, '-'))
print(df_1.dtypes) # 按列列出每列的数据类型
# A int64
# B datetime64[ns]
# C float64
# D category
# E object
# F object
# G int64
# dtype: object
print('index'.center(50, '-'))
print(df_1.index, type(df_1.index)) # Int64Index([0, 1, 2, 3], dtype='int64')
# Int64Index([0, 1, 2, 3], dtype='int64') <class 'pandas.core.indexes.numeric.Int64Index'>
print('columns'.center(50, '-'))
print(df_1.columns, type(df_1.columns)) # Index(['A', 'B', 'C', 'D', 'E', 'F'], dtype='object')
# Index(['A', 'B', 'C', 'D', 'E', 'F', 'G'], dtype='object') <class 'pandas.core.indexes.base.Index'>
print('values'.center(50, '-'))
print(df_1.values, type(df_1.values)) # 打印值,结果未numpy结构
# [[1 Timestamp('2019-09-30 00:00:00') 2.0 1 'a' 'beer' 1]
# [1 Timestamp('2019-09-30 00:00:00') 2.0 2 'b' 'beer' 5]
# [1 Timestamp('2019-09-30 00:00:00') 2.0 3 'c' 'beer' 4]
# [1 Timestamp('2019-09-30 00:00:00') 2.0 4 'd' 'beer' 4]] <class 'numpy.ndarray'>
print('describe'.center(50, '-'))
print(df_1.describe()) # 打印值,结果为numpy结构
# A C G
# count 4.0 4.0 4.000000 计数
# mean 1.0 2.0 2.500000 平均数
# std 0.0 0.0 1.290994 方差
# min 1.0 2.0 1.000000 最小值
# 25% 1.0 2.0 1.750000 第一分位数
# 50% 1.0 2.0 2.500000 第二分位数
# 75% 1.0 2.0 3.250000 第三分位数
# max 1.0 2.0 4.000000 最大值
print('数据翻转'.center(50, '-'))
print(df_1.T) # 打印值,结果未numpy结构
# 0 ... 3
# # A 1 ... 1
# # B 2019-09-30 00:00:00 ... 2019-09-30 00:00:00
# # C 2 ... 2
# # D 1 ... 4
# # E a ... d
# # F beer ... beer
# # G 1 ... 4
#
# [7 rows x 4 columns]
print('数据按照列排序'.center(50, '-'))
# axis=1,按照行进行排序
# axis=0,按照列进行排序
# ascending=False倒序输出,True为正序输出
print('按照index排序'.center(50, '-'))
print(df_1.sort_index(axis=0, ascending=False))
# A B C D E F G
# 3 1 2019-09-30 2.0 4 d beer 4
# 2 1 2019-09-30 2.0 3 c beer 4
# 1 1 2019-09-30 2.0 2 b beer 5
# 0 1 2019-09-30 2.0 1 a beer 1
print('G列排序'.center(50, '-'))
print(df_1.sort_values(by='G', ascending=False)) # 按照一列值排序
# A B C D E F G
# 1 1 2019-09-30 2.0 2 b beer 5
# 2 1 2019-09-30 2.0 3 c beer 4
# 3 1 2019-09-30 2.0 4 d beer 4
# 0 1 2019-09-30 2.0 1 a beer 1
print('G,D列排序'.center(50, '-'))
print(df_1.sort_values(by=['G', 'D'], ascending=False)) # 按照多列值排序
# A B C D E F G
# 1 1 2019-09-30 2.0 2 b beer 5
# 3 1 2019-09-30 2.0 4 d beer 4
# 2 1 2019-09-30 2.0 3 c beer 4
# 0 1 2019-09-30 2.0 1 a beer 1
print('数据按照行排序'.center(50, '-'))
index = list(range(4))
col = ['A', 'B', "C", 'D', 'E']
d_sort = pd.DataFrame(np.arange(20).reshape(4, 5), index=index, columns=col)
# print(d_sort)
print(d_sort.sort_values(by=[1, 2], axis=1, ascending=False)) # 按照行值排序
# E D C B A
# 0 4 3 2 1 0
# 1 9 8 7 6 5
# 2 14 13 12 11 10
# 3 19 18 17 16 15
5、选择数据
index = ['A', 'B', "C", 'D', 'E']
dates = pd.date_range('20191001', periods=10)
df = pd.DataFrame(np.random.randn(10, 5), index=dates, columns=index)
# print(df)
print('选择某列'.center(50, '-'))
print(df['A'])
# 2019-10-01 -0.595401
# 2019-10-02 1.264714
# 2019-10-03 1.179423
# 2019-10-04 -0.516471
# 2019-10-05 0.891850
# 2019-10-06 -0.011205
# 2019-10-07 -0.206089
# 2019-10-08 0.972745
# 2019-10-09 -0.135309
# 2019-10-10 1.590818
# Freq: D, Name: A, dtype: float64
print('切片选择'.center(50, '-'))
print(df[0:3]) # 按照行数切片
print(df['2019-10-02':'2019-10-05']) # 按照索引值进行切片
# A B C D E
# 2019-10-01 -0.595401 0.337930 0.034220 1.472752 -0.555414
# 2019-10-02 1.264714 0.518856 -1.148349 1.674159 -0.473919
# 2019-10-03 1.179423 2.036095 -0.719042 1.607909 2.659472
# A B C D E
# 2019-10-02 1.264714 0.518856 -1.148349 1.674159 -0.473919
# 2019-10-03 1.179423 2.036095 -0.719042 1.607909 2.659472
# 2019-10-04 -0.516471 1.733509 -0.177231 0.260795 -0.106666
# 2019-10-05 0.891850 0.665301 0.013627 -1.346193 0.222099
# 按照行切片[0:3]值切片到了0-2行
# 按照值'2019-10-02':'2019-10-05'切片,切到完整的日期范围
print('按照行精确选择'.center(50, '-'))
print(df.loc['2019-10-02', ['A', 'B']]) # 按照行精确选择列
# A 1.264714
# B 0.518856
# Name: 2019-10-02 00:00:00, dtype: float64
print('行号选择数据'.center(50, '-'))
print(df.iloc[3, 1])
# 1.7335085248615345
# 行数从0开始计数
print(df.iloc[3:5, 0:2]) # 输出4到到5行的数据,1到2列的数据
# A B
# 2019-10-04 -0.516471 1.733509
# 2019-10-05 0.891850 0.665301
# 切片从0开始计数,顾头部顾尾
print('混合选择'.center(50, '-'))
print(df.ix[0:3, ['B', 'C']])
# B C
# 2019-10-01 0.337930 0.034220
# 2019-10-02 0.518856 -1.148349
# 2019-10-03 2.036095 -0.719042
print('条件选择'.center(50, '-'))
print(df[df.A > 0])
# A B C D E
# 2019-10-01 0.391314 0.647378 0.065032 -0.436882 -0.482698
# 2019-10-02 1.742555 0.374014 0.737914 1.708461 0.328336
# 2019-10-03 0.024506 -0.455824 -0.397145 1.523103 1.361226
# 2019-10-04 0.140041 -0.604164 -0.397656 -0.423711 -0.626598
# 2019-10-05 0.027898 0.159293 -1.000558 0.921370 -1.613052
# 2019-10-08 1.411249 -1.292006 0.140944 0.699647 -0.065080
# 2019-10-10 0.306495 0.590515 -0.524972 0.521179 -0.805736