pandas之Series数据分析
python爬虫之数据分析包pandas
一.pandas介绍
pandas 是基于numpy构建的含有更高级数据结构和工具的数据分析包
类似于numpy的核心是ndarray,pandas也是围绕这series和datafrom > 两个核心数据结构
pandas的引入方式
1.安装方式
pip3 install pandas
2.引入方式
import pandas as pd
pandas的两大数据结构
1.Series
series是pandas的两种数据结构之一,可以理解为一维带标签数组
数组中的数据可以为任意类型(整数,字符串,浮点型,python objects等)
创建seriex
s = pd.Series(data, index=index)
"""
data 可以是list,array(数组),dictionary(字典)
NumPy 提供的 array() 函数直接将 Python 数组转换为 ndarray 数组,array() 接受一切序列类型的对象
"""
price = pd.Series([456,716,125])
price
"""
0 456
1 716
2 125
dtype:int64
"""
price = pd.Series([456,716,125], name='p')
price
"""
0 456
1 716
2 125
Name:p, dtype:int64
"""
p.mean()
"""
432.33333333333333
"""
p.sum()
"""
1297
"""
p.haed()
"""
price
0 456
1 716
Name:p, dtype:int64
"""
p.tail(2)
"""
1 716
2 125
Name:p, dtype:int64
"""
dic = {'three':100,'one':15,"two":78}
price = pd.Series(dic , name='p')
price
"""
three 100
one 15
two 78
Name: p, dtype: int64
"""
Series数据类型
price = pd.Series([1,2,3,4])
price.dtype
"""
dtype('int64')
"""
price = pd.Series([1,2,3,4.6])
price.dtype
"""
dtype('float64')
"""
ciyt = pd.Series(['wh','hz','sh','nj'])
city.dtype
"""
dtype('object')
"""
temp = pd.Series({},[],(1,2))
temp.dtype
"""
dtype('object')
"""
x = pd.Series(['2016-01-01','2017-01-01'])
x.dtype
"""
dtype('object')
"""
x = pd.Series(['a','b','a','c','d'],dtype='category')
x
"""
0 a
1 b
2 a
3 c
4 d
dtype:category
Categories(4,object):[a,b,c,d]
"""
boolean操作(布尔)
mask = pd.Series([True,True,False,True])
mask
"""
0 True
1 True
2 False
3 True
dtype:bool
"""
price[mask]
"""
0 1.0
1 2.0
3 4.0
dtype:float64
"""
mask2 = pd.Series([True,False,True,True])
nams|mask2
"""
有一个True就是True
0 True
1 True
2 True
3 True
"""
mask&mask3
"""
都True才为True
0 True
1 False
2 False
3 True
"""
~mask
"""
去反
0 False
1 False
2 True
3 False
""""
index操作
price
"""
0 1
1 2
2 3
3 4
"""
price[2]
"""
3
"""
price = pd.Series([1,2,3,4],index=['aa','bb','cc','dd'])
price
"""
aa 1
bb 2
cc 3
dd 4
"""
price.index
"""
index(['aa','bb','cc','dd'],dtype='object')
"""
日期相关
dates = pd.date_range('2019-01-01','2019-06-01',freq='M')
dates
"""
'M':每月最后一个日历
'W': 周
'D': 天
'H': 时
'T/min': 分
'S': 秒
DatetimeIndex(['2019-01-31', '2019-02-28', '2019-03-31', '2019-04-30',
'2019-05-31'],
dtype='datetime64[ns]', freq='M')
"""
tempature = pd.Series([10,11,20,27,29],index=dates)
tempature
"""
[五个值]
2019-01-31 10
2019-02-28 11
2019-03-31 20
2019-04-30 27
2019-05-31 29
Freq: M, dtype: int64
"""
切片
temp = pd.Series([12,14,15,18])
temp[0]
temp[2]
"""
12
15
"""
temp = pd.Series([12,14,15,18],index=['a','b','c','d'])
temp
temp['c']
"""
a 12
b 14
c 15
d 18
dtype:int64
15
"""
temp.iloc[2]
"""
15
"""
修改/增加/删除Series中的值
temp['a'] = 100
temp.iloc[1] = 200
temp
"""
修改
a 100
b 200
c 15
d 18
dtype:int64
"""
temp
统计函数summary,statistics
temp.min() # 最小
temp.sum() # 求和
temp.median() # 平均
temp.quantile(0.1)
temp.quantile(0.25)
temp.quantile(0.5)
"""
7.8
9.0
11.0
"""
temp.describe()
"""
count 2.000000
mean 11.000000
std 5.656854
min 7.000000
25% 9.000000
50% 11.000000
75% 13.000000
max 15.000000
dtype: float64
"""
temp=pd.Series(['hw','apple','vivo','mi','hw','oppo','samsung','vivo'],dtype='category')
temp.value_count()
"""
vivo 2
hw 1
samsung 1
oppo 1
mi 1
apple 1
dtype:int64
"""
向量化操作与广播
price = pd.Series([10,20,30,40], index=['o','t','t','t'])
price*2
"""
运算: + - * /
o 20
t 40
t 60
t 80
"""
price+100
"""
o 110
t 120
t 130
t 140
"""
s = pd.Series([10,20,30], index=[0,1,2])
s1 = pd.Series([40,50,60,70], index=[1,2,3,4])
s+s1
"""
NaN 在pandas中表示不是一个数字
0 NaN
1 60
2 80
3 NaN
4 NaN
"""
迭代iteration
for num i s:
print(num)
"""
1
2
3
"""
20 in s
"""
False
"""
20 in s.values
"""
True
"""
2 in s
"""
2 在index中
True
"""
for k,v in s.items():
print(k,v)
"""
0 10.0
1 20.0
2 30.0
"""